SET DEFAULT iterative logical name translation

[alanfe...@gmail.com]

This used to be broken. Strange things would happen under certain conditions, like only the directory being changed or the directory being hidden in sys$disk. It appears to be fixed on EISNER. Is there documentation of this fix somewhere? I've been searching on the web, only to come up empty. TIA! (^_^(

[Jan-Erik Söderholm]

Den 2021-10-22 kl. 08:45, skrev alanfe...@gmail.com:
> This used to be broken. Strange things would happen under certain conditions, like only the directory being changed or the directory being hidden in sys$disk. It appears to be fixed on EISNER. Is there documentation of this fix somewhere? I've been searching on the web, only to come up empty. TIA! (^_^(
>

What used to be brooken and what fix?
Care to post an example?

[alanfe...@gmail.com]

On Friday, October 22, 2021 at 5:58:22 AM UTC-4, Jan-Erik Söderholm wrote:

> Den 2021-10-22 kl. 08:45, skrev ala...@ g.com:
> > This used to be broken. Strange things would happen under certain conditions, like only the directory being changed or the directory being hidden in sys$disk. It appears to be fixed on EISNER. Is there documentation of this fix somewhere? I've been searching on the web, only to come up empty. TIA! (^_^(
> >
> What used to be brooken and what fix?
> Care to post an example?

I can't post an example, as I don't have access to a system where it was broken. But I have the following notes:

The DCL program SET DEFAULT has two problems with nested logical
names:

1.) If the first translation has a trailing colon and there is no
explicit directory-spec in the second translation, SYS$DISK is changed
to the 1st translation and the directory portion is not changed. The
"actual" current directory remains hidden in SYS$DISK.

2.) If the 1st translation has no trailing colon, then only the
directory portion of the default is changed! This can leave SYS$DISK
incorrect thereby leaving the default incorrect.

Additionally, in certain circumstances, SET DEFAULT returns an error
but still changes your process to a bad default.

I'd like to know what version of VMS this was fixed in and the ECO release notes if anyone can find them. There's an excellent post in cov from way, way back describing this, perhaps in a better way, but I can't find it.

[alanfe...@gmail.com]

I found the post! Here's an example with some context. I can post or reference the entire post if it's okay to do so. (Yeah, I'm a bit paranoid.)

Anyway, the details:

"As we all know", the default directory information is stored in two
pieces. The default device (stored as the equivalence name for the
logical name SYS$DISK:) and the default directory string (stored in
process P1 space named by PIO$GT_DDSTRING). Thus, changing the default
directory potentially involves changing two pieces of information.

The SET DEFAULT command looks at its given argument to determine
which pieces of data need to be changed. The problem is that it is
"stupid". If the given specification OR ITS FIRST TRANSLATION (if
a logical name) contains a colon, then SYS$DISK will be modified.
If no colon is found, ONLY THE DIRECTORY PATH will be changed.

So, let's look at the problem again.

$ SET DEFAULT HOME ! No colons here...

translation of HOME yeilds "SYS$LOGIN". no colons there either...

well, there's gotta be SOMETHING to do. Lets figure out
what directory path to store....

no directory path in SYS$LOGIN. translate it again.

"device:[directory]"

Ah ha! A directory path. Extract it and change the default
path to "[directory]"

All done!

So, there we have it. Isn't this brilliant? :) I run into this "gotcha"
every now and then when I don't remember to place a trailing colon on a
logical name, too. As far as I know, this behaviour isn't documented.
Apparently the VMS developers never make this "error", so never see the
problem. :)

[Lawrence D’Oliveiro]

On Friday, October 22, 2021 at 11:45:15 PM UTC+13, alanfe...@gmail.com wrote:
>
[intricacies of semantics of SET DEFAULT]

> So, there we have it. Isn't this brilliant? :) I run into this "gotcha"
> every now and then when I don't remember to place a trailing colon on a
> logical name, too. As far as I know, this behaviour isn't documented.
> Apparently the VMS developers never make this "error", so never see the
> problem. :)

Could be worse. Imagine, for the sake of argument, that somebody designed an OS where you could only use a single letter to name each disk device.

Crazy, I know. Nobody in their right mind would propose such a thing, would they? I mean, would you entrust mission-critical business functions to an OS that could only handle 26 drive letters?

[VAXman-]

ROTFLMFAO!

--
VAXman- A Bored Certified VMS Kernel Mode Hacker VAXman(at)TMESIS(dot)ORG

I speak to machines with the voice of humanity.

[Arne Vajhøj]

Probably nobody today.

Arne

[Phillip Helbig (undress to reply)]

In article <00B6ABCB...@SendSpamHere.ORG>, VAXman-

@SendSpamHere.ORG writes:

> > Could be worse. Imagine, for the sake of argument, that somebody
> > designed an OS where you could only use a single letter to name each
> > disk device.
> >
> > Crazy, I know. Nobody in their right mind would propose such a
> > thing, would they? I mean, would you entrust mission-critical business
> > functions to an OS that could only handle 26 drive letters?
>
> ROTFLMFAO!

I'm sure that when VAXman gets a chance to configure a Windows system,
he would try to use those nine letters for the drive names, but then
give up because two of them are repeated. :-)

Then just set up two drives: FO. :-D

[VAXman-]

In article <sl306k$4gp$1...@gioia.aioe.org>, hel...@asclothestro.multivax.de (Phillip Helbig (undress to reply)) writes:
>In article <00B6ABCB...@SendSpamHere.ORG>, VAXman-
>@SendSpamHere.ORG writes:
>

>> > Could be worse. Imagine, for the sake of argument, that somebody
>> > designed an OS where you could only use a single letter to name each
>> > disk device.
>> >
>> > Crazy, I know. Nobody in their right mind would propose such a
>> > thing, would they? I mean, would you entrust mission-critical business
>> > functions to an OS that could only handle 26 drive letters?
>>
>> ROTFLMFAO!
>

>I'm sure that when VAXman gets a chance to configure a Windows system,
>he would try to use those nine letters for the drive names, but then
>give up because two of them are repeated. :-)
>
>Then just set up two drives: FO. :-D

NOT that I would EVER configure a WEENDOZE box but if a were to, I'd use the
letters FJB.

[Bill Gunshannon]

On 10/23/21 7:19 PM, VAX...@SendSpamHere.ORG wrote:
> In article <2a79c8d5-1812-4fcb...@googlegroups.com>, =?UTF-8?Q?Lawrence_D=E2=80=99Oliveiro?= <lawren...@gmail.com> writes:
>> On Friday, October 22, 2021 at 11:45:15 PM UTC+13, alanfe...@gmail.com wrote:
>>>
>> [intricacies of semantics of SET DEFAULT]
>>> So, there we have it. Isn't this brilliant? :) I run into this "gotcha"
>>> every now and then when I don't remember to place a trailing colon on a
>>> logical name, too. As far as I know, this behaviour isn't documented.
>>> Apparently the VMS developers never make this "error", so never see the
>>> problem. :)
>>
>> Could be worse. Imagine, for the sake of argument, that somebody designed an OS where you could only use a single letter to name each disk device.
>>
>> Crazy, I know. Nobody in their right mind would propose such a thing, would they? I mean, would you entrust mission-critical business functions to an OS that could only handle 26 drive letters?
>
> ROTFLMFAO!
>

Why? This was at the same time when your only had 640K of
memory. And who would ever need more? Heck, why would
anyone ever need a computer in their house?

Hindsight is always 20/20.

bill

[Stephen Hoffman]

*You rang?*

Logical names are volatile key-value store design and straight out of
~1984, and long since and widely used and misused.

For device redirection, they're adequate.

Logical names have had "surprising" edge cases ~forever, too.

The whole of the default directory stuff cited is just utterly
hilarious, but that design is just too entrenched to ever be fixed and
the existing stuff deprecated and removed.

Search lists too are surprisingly fragile, particularly with missing
directories potentially in the mix.

I'd wager that most existing apps don't specify the filenames as would
be preferable, and many OpenVMS developers aren't aware of the
"preferred" filename defaulting and translation sequence. Or aren't
interested in using that ordering.

And then there's what I'd consider the misuse of logicals as a volatile
and out-of-band and poorly implemented management UI. q.v. CRTL.

Alas, VSI is unlikely to be interested in adding additional checks as
those sorts of changes have a history of breaking existing and ~working
configurations, too. q.v. Hyrum's Law.

Best case, app preferences support become available within OpenVMS,
preferably non-volatile, and existing app-specific file and device
accesses can then all be redirected from within the app-associated
preferences, as permitted by local security policies.

But this current logical name design? I'm not sure it's even
particularly salvageable short of the V3-to-V4-scale rework of the
whole logical names design, if it's even worth salvaging, and that
salvaging necessarily preserving most of the edge cases for existing
consumers. The original logical names were a great idea within the
limits of the PDP-11 / RSX-11M era. The V4 stuff was far more flexible
and with better APIs, but whether it really improved this whole area?
Nowadays, past isolated device or isolated file redirection—where apps
have been created to allow that—the whole of the design is really
limited and showing its age.





--
Pure Personal Opinion | HoffmanLabs LLC

[alanfe...@gmail.com]

On Sunday, October 24, 2021 at 12:22:44 PM UTC-4, Stephen Hoffman wrote:
> On 2021-10-22 06:45:02 +0000, alanfe...@gmail.com said:
>
> > This used to be broken. Strange things would happen under certain
> > conditions, like only the directory being changed or the directory
> > being hidden in sys$disk. It appears to be fixed on EISNER. Is there
> > documentation of this fix somewhere? I've been searching on the web,
> > only to come up empty. TIA! (^_^(
> *You rang?*

Yep! That was the best part of the show, no? When Lurch would suddenly turn into the scene seemingly out of nowhere and say, "You rang?"
[...]

>
> Logical names are volatile key-value store design and straight out of
> ~1984, and long since and widely used and misused.
>
> For device redirection, they're adequate.
>
> Logical names have had "surprising" edge cases ~forever, too.
>
> The whole of the default directory stuff cited is just utterly
> hilarious, but that design is just too entrenched to ever be fixed and
> the existing stuff deprecated and removed.

But Hoff, it _was_ fixed! Looks like no one knows when or what version of VMS it was fixed in.

Here's a bonus question. When was EDT updated to allow more than 22 lines? IDK, but it's made me happy!

Bonus No. 2: In what version of VMS was "Proactive Reclamation of Memory from Idle Processes" introduced? This one I found on my own. Turns out it was v5.4-3. Made a big difference in my main system at the time!
>
[...]

> --
> Pure Personal Opinion | HoffmanLabs LLC

AEF

[Jan-Erik Söderholm]

Den 2021-10-25 kl. 06:28, skrev alanfe...@gmail.com:
> On Sunday, October 24, 2021 at 12:22:44 PM UTC-4, Stephen Hoffman wrote:
>> On 2021-10-22 06:45:02 +0000, alanfe...@gmail.com said:
>>
>>> This used to be broken. Strange things would happen under certain
>>> conditions, like only the directory being changed or the directory
>>> being hidden in sys$disk. It appears to be fixed on EISNER. Is there
>>> documentation of this fix somewhere? I've been searching on the web,
>>> only to come up empty. TIA! (^_^(
>> *You rang?*
>
> Yep! That was the best part of the show, no? When Lurch would suddenly turn into the scene seemingly out of nowhere and say, "You rang?"
> [...]
>>
>> Logical names are volatile key-value store design and straight out of
>> ~1984, and long since and widely used and misused.
>>
>> For device redirection, they're adequate.
>>
>> Logical names have had "surprising" edge cases ~forever, too.
>>
>> The whole of the default directory stuff cited is just utterly
>> hilarious, but that design is just too entrenched to ever be fixed and
>> the existing stuff deprecated and removed.
>
> But Hoff, it _was_ fixed! Looks like no one knows when or what version of VMS it was fixed in.
>
> Here's a bonus question. When was EDT updated to allow more than 22 lines? IDK, but it's made me happy!

There as an EDT patch just a year ago or so from VSI.
Had something with numnber of lines to do, I think...

[Simon Clubley]

On 2021-10-25, alanfe...@gmail.com <alanfe...@gmail.com> wrote:
>
> Here's a bonus question. When was EDT updated to allow more than 22 lines? IDK, but it's made me happy!
>

$ set response/mode=good_natured

It was a procrastination project by someone at VSI. :-)

(IOW, a project you work on when you are stuck on what you are _really_
supposed to be working on.)

Simon.

--
Simon Clubley, clubley@remove_me.eisner.decus.org-Earth.UFP
Walking destinations on a map are further away than they appear.

[Stephen Hoffman]

On 2021-10-25 04:28:45 +0000, alanfe...@gmail.com said:

> But Hoff, it _was_ fixed! Looks like no one knows when or what version
> of VMS it was fixed in.

If this purported logical name colon-oscopy makes you happy and if
logical names do what you need, by all means enjoy and use logical
names. They are what is available, of course.

For my usage, logical names are a weak and volatile system-integrated
key-value store design straight from the 1980s, and largely intended
for app customization, and competitively woefully inadequate for app
needs past device and file redirection, and I'm increasingly skeptical
there.

Logical names provide one facet of app customization semi-adequately,
and that is not an issue I tend to have encounter or create in newer
app designs, while poorly providing for other app customization
requirements, and the whole of logical-name-based settings maintenance
is just utterly and stupidly disconnected from the apps themselves.

The whole realm of logical names and app customization needs an
overhaul of a scale well past the VAX/VMS V4.0 overhaul.

> Here's a bonus question. When was EDT updated to allow more than 22
> lines? IDK, but it's made me happy!

EDT was long ago considered deprecated and ~immutable, and now it's
apparently not. That's a decision that VSI gets to make, of course.
Whether its apparent commutation was a wise decision remains to be
seen. But then I tend to use EDT only when that's the last editor
available this side of PATCH. For those that do still prefer to use
EDT, yay, it's not dead yet. Maybe some of the 🤬 in the EDT help gets
fixed too, as VSI develops their strategy for text editor support.

BTW: LSEDIT permits the EDT keypad, and has supported far larger
displays, and with app development features massively better than EDT.

[alanfe...@gmail.com]

On Monday, October 25, 2021 at 12:31:01 PM UTC-4, Stephen Hoffman wrote:

"Logical name colonoscopy?" I was talking about SET DEFAULT and my program (DCL command procedure) TO.COM. Yes, TO.COM does stuff with logical names. But it does a lot more than that. SET DEFAULT had 2 or 3 silly bugs that I figured wouldn't take long to fix. But year after year they didn't fix it. This is one of the reasons I wrote TO.COM. Hey, at least one person has used it, as he wrote to me about it! I figured if anyone would ever use it I better give it lots of features and make it as bulletproof as possible.

I get the impression you don't think highly of logical names! I can't imagine removing them from VMS though. Sounds like a lot of work, as they are so ubiquitous. And there are all the 3rd-party apps that use them!

Anyway, I am curious about when, what version, and want to get a copy of the release notes for it. I found them for EDT now having more than 22 lines! (You also don't seem to think much of EDT!)

[The following is a tad long, but you've now and in the past told me there are better editing tools than EDT, so I'd like to tell you why I didn't use them much.]

Speaking of EDT, I have to say it has served me well. When I was at a remote place to do an experiment in the late 1980s, I tried EVE/TPU/whatever. Didn't work. Something with terminal settings -- maybe is was the Eightbit thing. I don't recall for sure. OTOH, I found that EDT worked everywhere under any circumstances without any fuss! "It just worked." But its 22-line limit became really annoying when bigger screens became available. And it was finally extended last year! Another project that probably took very little time to implement. But no. We were stuck with the 22-line limit and the silly SET DEFAULT bugs year after year after year for no good reason.

[Stay with me here. I get to the editors shortly.] At one job c. 1999, I was working on exporting data from a database -- using DCL, no less! As a former mere physicist, I was excited that I figured out a way to do it. And I only had DCL to work with. Sure, tell me all the better ways I could and should have done it. (Actually, maybe not.) All I had was DCL. Oh, and I had to export ALLIN1 Word Perfect documents, too! A third party vendor was going to charge us something like $11,000 plus expenses _just_ for the database export! I saved my company at least that much money. Now, I mention this because I was very frustrated that EVE didn't have EDT's SET NOTRUNCATE feature. I sorely needed that for a particular task but EDT didn't support long-enough lines. So I just had to struggle. (There was no native support in that database for exporting data that anyone could find, including the third party we hired to put the exported data into their newfangled database.)

I didn't use VMS from 2010 until 2015 when (as you have probably already guessed, the migrate-off-of-VMS part), I was to help export the file trading system to a Unix system with Autosys for the scheduling. VMS was pretty soon a very minor part of the process until we finally found the contact for two very important files they sent us every workday. Maybe it was late 2016 or early 2017. In 2018 I left the company. None of this was app development, unless you count the export work c. 1999, where EDT and EVE each had their respective annoying shortcoming.

Out of curiosity I recently looked at the freeware page at VSI and saw an old version of TO.COM (v4.3.1) and that the link to download it was broken. I did the "Contact Us" routine and was told it could be fixed and would I like to submit an updated version. "Sure!" I said.

So I was really glad to see EDT now has unlimited lines. Better late than never! Actually, it's kind of mind-blowing to see EDT on my screen with roughly 50 lines and sometimes a little more!

As for EVE and LSE, I never had any luck with those programs. EVE didn't work at remote labs without a lot of fuss. Locally I tried to customize EVE and at one point it kept saying "Compilation aborted at line 1." But EDT always worked and worked everywhere. Once I learned the nokeypad commands, I customized it to my liking. I even fiddled with it a little now. I already know it. I don't already know the others. Modifying a key definition was super easy. You could see them all at once, and just edit the one you needed to modify, and close the file! Oh, and I forgot what the problem with LSEDIT was. EVE and LSE were too-powerful tools for my needs. Like using a blow torch to light a cigarette.

As for LSEDIT supporting larger displays, I don't seem to need that. As for app-development features, I don't need them either. I'm pretty much only working on TO.COM. Maybe I'll spruce up a couple of others. If I ever get an OpenVMS job again, maybe I'll check them out again.

To end on a positive note, I was very happy when "Proactive Reclamation of Memory from Idle Processes" came on the scene. And it was v5.4-3. Boy that made things run better! At my last job, before my app died, we had it running on a fleet of as many as 40 MicroVAXes at one point. One of them in London had only 8 MB of RAM, while the others running the same app had 40 MB. Memory was very tight, but it still ran like a champ!

Hey, a second positive note! My favorite feature of EVE is the box editing. I think I can probably make use of it to improve my help page, which has two columns in it. And I did occasionally make use of it in the past.

Thanks for your reply.

[Dave Froble]

On 10/25/2021 12:30 PM, Stephen Hoffman wrote:
> On 2021-10-25 04:28:45 +0000, alanfe...@gmail.com said:
>
>> But Hoff, it _was_ fixed! Looks like no one knows when or what
>> version of VMS it was fixed in.
>
> If this purported logical name colon-oscopy makes you happy and if
> logical names do what you need, by all means enjoy and use logical
> names. They are what is available, of course.

I do use them, to some extent. I don't go in for some of the quirky
things, such as rooted logicals. I don't like things to be hidden.

I'm quite happy with the way I use logicals.
I am not guilty of how others may use them.

> For my usage, logical names are a weak and volatile system-integrated
> key-value store design straight from the 1980s, and largely intended
> for app customization, and competitively woefully inadequate for app
> needs past device and file redirection, and I'm increasingly
> skeptical there.

WHOA!!!

Where did that "app customization" thing come from?

Device and file redirection (your terminology) is a good thing, and
logicals used in a good manner do that quite well.

> Logical names provide one facet of app customization semi-adequately,
> and that is not an issue I tend to have encounter or create in newer
> app designs, while poorly providing for other app customization
> requirements, and the whole of logical-name-based settings
> maintenance is just utterly and stupidly disconnected from the apps
> themselves.

Got to agree, since logicals have no place in the apps.

> The whole realm of logical names and app customization needs an
> overhaul of a scale well past the VAX/VMS V4.0 overhaul.

The concept needs to die ...


--
David Froble Tel: 724-529-0450
Dave Froble Enterprises, Inc. E-Mail: da...@tsoft-inc.com
DFE Ultralights, Inc.
170 Grimplin Road
Vanderbilt, PA 15486

[Dave Froble]

On 10/25/2021 2:28 PM, alanfe...@gmail.com wrote:

> But it does a lot more than that. SET DEFAULT had 2 or 3 silly bugs that I figured wouldn't take long to fix.

What bugs? I've never seen any.

[Phillip Helbig (undress to reply)]

In article <sl6m42$4v0$1...@dont-email.me>, Stephen Hoffman

<seao...@hoffmanlabs.invalid> writes:

> For my usage, logical names are a weak and volatile system-integrated
> key-value store design straight from the 1980s, and largely intended
> for app customization, and competitively woefully inadequate for app
> needs past device and file redirection, and I'm increasingly skeptical
> there.

Cluster-wide logical names, especially those visible only to a certain
group (but cluster-wide) are very nice. Does linux have something
similar (and I don't mean turd files).

> EDT was long ago considered deprecated and ~immutable, and now it's
> apparently not.

Much faster than TPU: no cursor slow motion, doesn't read the whole 1-GB
file if it just needs to change the top few lines. Probably easier to
script and run in batch.

> BTW: LSEDIT permits the EDT keypad, and has supported far larger
> displays, and with app development features massively better than EDT.

TPU supports the EDT keypad. So does emacs, for that matter. But there
is much more to EDT than the keypad. It's a way of life. :-)

[Arne Vajhøj]

On 10/25/2021 3:54 PM, Phillip Helbig (undress to reply) wrote:
> In article <sl6m42$4v0$1...@dont-email.me>, Stephen Hoffman
> <seao...@hoffmanlabs.invalid> writes:
>> EDT was long ago considered deprecated and ~immutable, and now it's
>> apparently not.
>
> Much faster than TPU: no cursor slow motion, doesn't read the whole 1-GB
> file if it just needs to change the top few lines. Probably easier to
> script and run in batch.

EVE/TPU is actually very powerful for scripting.

Arne

[Stephen Hoffman]

On 2021-10-25 19:54:43 +0000, Phillip Helbig (undress to reply said:

> In article <sl6m42$4v0$1...@dont-email.me>, Stephen Hoffman
> <seao...@hoffmanlabs.invalid> writes:
>
>> For my usage, logical names are a weak and volatile system-integrated
>> key-value store design straight from the 1980s, and largely intended
>> for app customization, and competitively woefully inadequate for app
>> needs past device and file redirection, and I'm increasingly skeptical
>> there.
>
> Cluster-wide logical names, especially those visible only to a certain
> group (but cluster-wide) are very nice. Does linux have something
> similar (and I don't mean turd files).

Yes; there are more flexible and more capable alternatives, as has been
discussed on each of the previous times you've asked this question.

The OpenVMS APIs available for apps to access a distributed directory
and to access and maintain settings are less well developed than those
on other platforms, which is the crux of my comments.

>> EDT was long ago considered deprecated and ~immutable, and now it's
>> apparently not.
>
> Much faster than TPU: no cursor slow motion, doesn't read the whole
> 1-GB file if it just needs to change the top few lines. Probably
> easier to script and run in batch.

Main memory on an Alpha runs at a fraction of the speed of main storage
on recent-generation client hardware, and recent-generation client main
memory surpasses Alpha processor cache speeds.

(Alpha EV7 L1 cache bandwidth: 7.77 GBps, versus recent-gen 204 GBps to
main memory. Alpha local main memory bandwidth 4.6 GBps, versus
recent-gen 7.4 GBps SSD read bandwidth. Etc.)

On OpenVMS, Eve and particularly TPU were built to script.

Most editors now are scriptable, of course. Just ask an emacs user. Or
a teco user.

And dedicated and portable scripting languages widely available, not
the least of which are Perl, Python, and Lua.

>> BTW: LSEDIT permits the EDT keypad, and has supported far larger
>> displays, and with app development features massively better than EDT.
>
> TPU supports the EDT keypad. So does emacs, for that matter. But
> there is much more to EDT than the keypad. It's a way of life. :-)

Until and unless your preferred 1980s-era tooling ceases to work for
you and your needs, I would expect nothing less.

[Stephen Hoffman]

On 2021-10-25 19:32:15 +0000, Dave Froble said:

> On 10/25/2021 12:30 PM, Stephen Hoffman wrote:
>
>> For my usage, logical names are a weak and volatile system-integrated
>> key-value store design straight from the 1980s, and largely intended
>> for app customization, and competitively woefully inadequate for app
>> needs past device and file redirection, and I'm increasingly skeptical
>> there.
>
> WHOA!!!
>
> Where did that "app customization" thing come from?
>
> Device and file redirection (your terminology) is a good thing, and
> logicals used in a good manner do that quite well.

Logical names are a central part of the available app-customization API
available within OpenVMS.

The whole point of logical names is to customize the default behaviors
of and the file and device references of apps, and of OpenVMS itself.

Logical names work decently well for file and device redirections, and
particularly when the developers know about and use FNA/DNA/RLF.

Logical names tend to get messy with other app-specific customization
settings; when not re-directing file or device references.

A whole lot of the "magic" of logical names is tied to the proper
filename specification on the file opens too, and that usage is far
from ubiquitous among developers.

A volatile key-value store is both limited and limiting too, as app
customization APIs go.

And as I've mentioned up-thread, I'm finding it rather less common to
be redirecting individual files within apps. OpenVMS itself and its
archaic clustering "management interface", yes. But App file
references, not so much.

[Bill Gunshannon]

So is EMACS... :-)

bill

[alanfe...@gmail.com]

On Monday, October 25, 2021 at 3:47:23 PM UTC-4, Dave Froble wrote:
> On 10/25/2021 2:28 PM, alanfe...@gmail.com wrote:
>
> > But it does a lot more than that. SET DEFAULT had 2 or 3 silly bugs that I figured wouldn't take long to fix.
> What bugs? I've never seen any.

I posted on earlier in the thread. I think it's even mentioned in the docs once, but I don't recall where and doubt I can find it.

Here is my summary of them:

The DCL program SET DEFAULT has two problems with nested logical
names:


1.) If the first translation has a trailing colon and there is no
explicit directory-spec in the second translation, SYS$DISK is changed
to the 1st translation and the directory portion is not changed. The
"actual" current directory remains hidden in SYS$DISK.

[ I assume this mean if you specify <LNM>:[directory] ]

2.) If the 1st translation has no trailing colon, then only the
directory portion of the default is changed! This can leave SYS$DISK
incorrect thereby leaving the default incorrect.

[This is the example posted by someone long ago which I put in one of my first posts in this thread.]

Additionally, in certain circumstances, SET DEFAULT returns an error
but still changes your process to a bad default.

[Unfortunately, I don't recall any examples of this.]

> --
> David Froble Tel: 724-529-0450
> Dave Froble Enterprises, Inc. E-Mail: da...@tsoft-inc.com
> DFE Ultralights, Inc.
> 170 Grimplin Road
> Vanderbilt, PA 15486

AEF

[Lawrence D’Oliveiro]

On Monday, October 25, 2021 at 5:28:47 PM UTC+13, alanfe...@gmail.com wrote:
> When was EDT updated to allow more than 22 lines? IDK, but it's made me happy!

“EDT” and “happy” were never words I considered using in the same sentence ...

[Lawrence D’Oliveiro]

On Tuesday, October 26, 2021 at 7:28:13 AM UTC+13, alanfe...@gmail.com wrote:
> ... I tried EVE/TPU/whatever. Didn't work.

After suffering the loss of TECO and putting up with EDT, TPU was a welcome relief. I started with the source code for EVE, and customized it heavily. I called my version “PEEVE”.

[Dave Froble]

When I need to change "ABC" to "XYZ", EDT is very useful ...

[Dave Froble]

On 10/25/2021 9:25 PM, alanfe...@gmail.com wrote:
> On Monday, October 25, 2021 at 3:47:23 PM UTC-4, Dave Froble wrote:
>> On 10/25/2021 2:28 PM, alanfe...@gmail.com wrote:
>>
>>> But it does a lot more than that. SET DEFAULT had 2 or 3 silly bugs that I figured wouldn't take long to fix.
>> What bugs? I've never seen any.
>
> I posted on earlier in the thread. I think it's even mentioned in the docs once, but I don't recall where and doubt I can find it.
>
> Here is my summary of them:
>
> The DCL program SET DEFAULT has two problems with nested logical
> names:

Simple solution, do not use nested logical names ...

If you plan on pushing boundaries, expect to find boundaries ...

I tend to use the KISS principal ...

[Lawrence D’Oliveiro]

On Monday, October 25, 2021 at 2:30:30 AM UTC+13, Bill Gunshannon wrote:
> This was at the same time when your only had 640K of
> memory.

Both limits are just as old. But 640K was fixed ... eventually ... slowly ... painfully.

Whereas Microsoft still seem to believe that 26 drive letters ought to be enough for anybody.

[Henry Crun]

On 26/10/2021 8:49, Dave Froble wrote:
> On 10/25/2021 4:33 PM, Arne Vajhøj wrote:
>> On 10/25/2021 3:54 PM, Phillip Helbig (undress to reply) wrote:
>>> In article <sl6m42$4v0$1...@dont-email.me>, Stephen Hoffman
>>> <seao...@hoffmanlabs.invalid> writes:
>>>> EDT was long ago considered deprecated and ~immutable, and now it's
>>>> apparently not.
>>>
>>> Much faster than TPU: no cursor slow motion, doesn't read the whole 1-GB
>>> file if it just needs to change the top few lines.  Probably easier to
>>> script and run in batch.
>>
>> EVE/TPU is actually very powerful for scripting.
>>
>> Arne
>>
>
> When I need to change "ABC" to "XYZ", EDT is very useful ...
>

But when you want to replace "ABC" with "XYZ" in every line containing "UVW" you need TPU "Learn" capability

--
Mike R.
Home: http://alpha.mike-r.com/
QOTD: http://alpha.mike-r.com/qotd.php
No Micro$oft products were used in the URLs above, or in preparing this message.
Recommended reading: http://www.catb.org/~esr/faqs/smart-questions.html#before
and: http://alpha.mike-r.com/jargon/T/top-post.html
Missile address: N31.7624/E34.9691

[alanfe...@gmail.com]

On Tuesday, October 26, 2021 at 4:52:38 AM UTC-4, Henry Crun wrote:
> On 26/10/2021 8:49, Dave Froble wrote:
> > On 10/25/2021 4:33 PM, Arne Vajhøj wrote:
> >> On 10/25/2021 3:54 PM, Phillip Helbig (undress to reply) wrote:
> >>> In article <sl6m42$4v0$1...@dont-email.me>, Stephen Hoffman
> >>> <seao...@hoffmanlabs.invalid> writes:
> >>>> EDT was long ago considered deprecated and ~immutable, and now it's
> >>>> apparently not.
> >>>
> >>> Much faster than TPU: no cursor slow motion, doesn't read the whole 1-GB
> >>> file if it just needs to change the top few lines. Probably easier to
> >>> script and run in batch.
> >>
> >> EVE/TPU is actually very powerful for scripting.
> >>
> >> Arne
> >>
> >
> > When I need to change "ABC" to "XYZ", EDT is very useful ...
> >
> But when you want to replace "ABC" with "XYZ" in every line containing "UVW" you need TPU "Learn" capability
>

Nope. EDT can do it! (ED is EDIT/EDT/.)

$ ED B.B
1 ABC
*TYPE WH
1 ABC
2 ABC UVW
3 ABC
4 ABC UVW
5 ABC UVW
6 ABC
[EOB]
*S/ABC/XYZ/ ALL 'UVW'
2 XYZ UVW
4 XYZ UVW
5 XYZ UVW
3 substitutions
*TYPE WH
1 ABC
2 XYZ UVW
3 ABC
4 XYZ UVW
5 XYZ UVW
6 ABC
[EOB]
*

> --
> Mike R.
> Home: http://alpha.mike-r.com/
> QOTD: http://alpha.mike-r.com/qotd.php
> No Micro$oft products were used in the URLs above, or in preparing this message.
> Recommended reading: http://www.catb.org/~esr/faqs/smart-questions.html#before
> and: http://alpha.mike-r.com/jargon/T/top-post.html
> Missile address: N31.7624/E34.9691

AEF

[alanfe...@gmail.com]

On Tuesday, October 26, 2021 at 1:59:06 AM UTC-4, Dave Froble wrote:
> On 10/25/2021 9:25 PM, alanfe...@gmail.com wrote:
> > On Monday, October 25, 2021 at 3:47:23 PM UTC-4, Dave Froble wrote:
> >> On 10/25/2021 2:28 PM, alanfe...@gmail.com wrote:
> >>
> >>> But it does a lot more than that. SET DEFAULT had 2 or 3 silly bugs that I figured wouldn't take long to fix.
> >> What bugs? I've never seen any.
> >
> > I posted on earlier in the thread. I think it's even mentioned in the docs once, but I don't recall where and doubt I can find it.
> >
> > Here is my summary of them:
> >
> > The DCL program SET DEFAULT has two problems with nested logical
> > names:
> Simple solution, do not use nested logical names ...

Sure. Just don't use SYS$MANAGER and the like. Right. Actually the problem doesn't affect all nested logical names. No matter. They fixed it anyway. Can't get simpler than that!

Better yet, just get TO.COM! It has way more than just dealing with nested logical names. I hope to release the latest version sometime in next couple of weeks or so. (The link to the version presently online is broken.)

>
> If you plan on pushing boundaries, expect to find boundaries ...

???

[alanfe...@gmail.com]

I understand. Use the right tool for the job, I say. For me that is usually EDT.

One thing I really like about EDT is the key definitions. You can see them and change them without needing to recompile and make a new huge TPU section file (IIRC the TPU drill). Say you take a long vacation. You come back and can't remember some of your key definitions. What are you going to do? In EDT you can look them up very quickly. I don't recall such a feature for EVE. But that was many years ago.

One thing I really like about EVE is box editing. So if I need that, I use EVE. The learn feature is pretty cool, but again, you can't see what it is if you forget later. Certainly good for a repetitive one-shot task though. Or you can write down what it is!

If EVE or LSEDIT some other editor work for you, great! For my purposes it's like using a blow torch to light a cigarette. But some of you need a blow torch. That's fine with me. You've got it!

[alanfe...@gmail.com]

On Monday, October 25, 2021 at 9:25:10 PM UTC-4, alanfe...@gmail.com wrote:
> On Monday, October 25, 2021 at 3:47:23 PM UTC-4, Dave Froble wrote:
> > On 10/25/2021 2:28 PM, alanfe...@gmail.com wrote:
> >
> > > But it does a lot more than that. SET DEFAULT had 2 or 3 silly bugs that I figured wouldn't take long to fix.
> > What bugs? I've never seen any.
> I posted on earlier in the thread. I think it's even mentioned in the docs once, but I don't recall where and doubt I can find it.

I now seem to recall it was mentioned in one of the ECO release notes. It think the notes also said it would be fixed in a future version of VMS. Well, they were right!

[...]

> > --
> > David Froble Tel: 724-529-0450
> > Dave Froble Enterprises, Inc. E-Mail: da...@tsoft-inc.com
> > DFE Ultralights, Inc.
> > 170 Grimplin Road
> > Vanderbilt, PA 15486
> AEF

AEF

[Arne Vajhøj]

On 10/26/2021 5:49 AM, alanfe...@gmail.com wrote:
> On Tuesday, October 26, 2021 at 4:52:38 AM UTC-4, Henry Crun wrote:
>> On 26/10/2021 8:49, Dave Froble wrote:
>>> On 10/25/2021 4:33 PM, Arne Vajhøj wrote:
>>>> On 10/25/2021 3:54 PM, Phillip Helbig (undress to reply) wrote:
>>>>> In article <sl6m42$4v0$1...@dont-email.me>, Stephen Hoffman
>>>>> <seao...@hoffmanlabs.invalid> writes:
>>>>>> EDT was long ago considered deprecated and ~immutable, and now it's
>>>>>> apparently not.
>>>>>
>>>>> Much faster than TPU: no cursor slow motion, doesn't read the whole 1-GB
>>>>> file if it just needs to change the top few lines. Probably easier to
>>>>> script and run in batch.
>>>>
>>>> EVE/TPU is actually very powerful for scripting.
>>>

Ha. We got an EDT wizard among us.

:-)

But still TPU is really a fullblown programming
language with procedures, if statement, loop
statement etc., so I will maintain the original
claim that it is good for scripting.

Arne

[Arne Vajhøj]

The English language decided on 26 letters.

The last Windows server version that required a unique drive letter per
disk was NT 4.0 Server (released 1996, end of support 2002/2004). All
newer Windows versions (2003, 2008, 2008R2, 2012, 2012R2, 2016, 2019
and 2022) supports volume mount points.

Arne

[Bill Gunshannon]

And, as is usually the case when people look back on the way things
used to be they target one thing while missing all of the contributing
factors. When there were only 26 drive letters (24 if you accept that
two were owned by floppies) how many physical drive could one actually
attach to a PC system?

bill

[Arne Vajhøj]

I don't even think it is a real problem today.

For systems with lots of disk space I believe that the normal model
is that the storage system present a few huge logical disks to
the system and the storage system translate that to a large number
of physical drives using RAID 10 or RAID 6 or whatever.

Arne

[Phillip Helbig (undress to reply)]

In article <61780458$0$697$1472...@news.sunsite.dk>,

=?UTF-8?Q?Arne_Vajh=c3=b8j?= <ar...@vajhoej.dk> writes:

> >>>> EVE/TPU is actually very powerful for scripting.
> >>>

> >>> When I need to change "ABC" to "XYZ", EDT is very useful ...
> >>>
> >> But when you want to replace "ABC" with "XYZ" in every line
> >> containing "UVW" you need TPU "Learn" capability
> >
> > Nope. EDT can do it! (ED is EDIT/EDT/.)

> Ha. We got an EDT wizard among us.
>
> :-)

Nope; that is bread and butter for many EDT users.

> But still TPU is really a fullblown programming
> language with procedures, if statement, loop
> statement etc., so I will maintain the original
> claim that it is good for scripting.

It depends on what one's needs are: right tool for the job and so on.

[Dave Froble]

On 10/26/2021 5:49 AM, alanfe...@gmail.com wrote:

Very nice !!!

That specific capability is something that I've personally never run into.

When working on a program, which is my main usage of editors, I'm very careful
of what I'm doing. Mass substitutions can drive a stake through your heart
if not very specific about what's happening. Thus, I'm not a user of some advanced,
and even normal, capabilities.

[Simon Clubley]

On 2021-10-25, Phillip Helbig (undress to reply) <hel...@asclothestro.multivax.de> wrote:
>
> TPU supports the EDT keypad. So does emacs, for that matter. But there
> is much more to EDT than the keypad. It's a way of life. :-)
>

Sure, provided you also use Oxen to plough your fields and use candles
to light your buildings.

EDT is the kind of editor the Amish would use. :-)

Simon.

--
Simon Clubley, clubley@remove_me.eisner.decus.org-Earth.UFP
Walking destinations on a map are further away than they appear.

[alanfe...@gmail.com]

I agree 100%. Blind mass substitution can be a disaster, esp. for short strings! I'm always careful when doing such things. Also, I don't think I ever needed this particular capability, but I know it exists, so I had to post it.

One day at a remote lab, our experiment was done. As a lowly grad student, it was my job to make backup copies all the 9-track tapes we wrote our data to. Don't want to lose all that hard-gained data!!! We had the lab to ourselves for several days or maybe even a week or week and a half. Don't want all that long hard work to go to waste!!!

Took about 10 min. to make a copy of each tape. So I thought I'd take the opportunity to catch up to everyone else on VMS. During each 10 min. interval I literally plowed through the DCL dictionary, and probably the User's manual and the EDT manual, too.

Again to all: If EDT is not up to your needs, sure, use whatever is. But those editors only get in my way. I am working on DCL scripts. Nothing terribly fancy. For this, EDT serves me well. All I am doing is sprucing up TO.COM (fixed a bug, too!) to resubmit the current version (a somewhat old one is there). And maybe check on other things.

My favorite thing about EDT is probably the way it does the key definitions.

Again, if anyone can find the release notes for fixing, or even the one describing the SET DEFAULT problem, please do post. (Was most likely the 6.0, 6.1, or 6.2 version of OpenVMS where the problem was described.)

> --
> David Froble Tel: 724-529-0450
> Dave Froble Enterprises, Inc. E-Mail: da...@tsoft-inc.com
> DFE Ultralights, Inc.
> 170 Grimplin Road
> Vanderbilt, PA 15486

AEF

[Lawrence D’Oliveiro]

On Tuesday, October 26, 2021 at 11:19:27 PM UTC+13, alanfe...@gmail.com wrote:
> On Monday, October 25, 2021 at 10:48:26 PM UTC-4, Lawrence D’Oliveiro wrote:
>> On Monday, October 25, 2021 at 5:28:47 PM UTC+13, alanfe...@gmail.com wrote:
>>> When was EDT updated to allow more than 22 lines? IDK, but it's made me happy!
>> “EDT” and “happy” were never words I considered using in the same sentence ...
> I understand. Use the right tool for the job, I say. For me that is usually EDT.

What drove me mad about EDT was its insistence on a “current direction” mode. Instead of having separate movement commands for going forward and backward by word, line etc, it had a single command for each, with separate “forward” and “backward” keys for setting the direction.

But since I could never remember what was the last direction key I hit, I would always have to hit it again, so every movement command became *two* keystrokes instead of one.

> One thing I really like about EDT is the key definitions. You can see them and change

> them without needing to recompile and make a new huge TPU section file ...

Emacs does all that, and more. It also has the advantage of being built on LISP, which is still one of the most advanced programming languages around.

[Lawrence D’Oliveiro]

On Wednesday, October 27, 2021 at 3:07:08 AM UTC+13, Bill Gunshannon wrote:
> When there were only 26 drive letters (24 if you accept that
> two were owned by floppies) how many physical drive could one actually
> attach to a PC system?

Your point was valid in 1980, and starting to seem a bit weak by 1990. Consider how it looks today.

I wonder how Dave Cutler feels, knowing that his next-generation “New Technology” OS, ostensibly created from the ground up to take full advantage of 32-bit hardware, is still hobbled by a limitation baked into its heart that dates back to the 8-bit era.

At least VMS never had to suffer that ignominy...

[Lawrence D’Oliveiro]

On Wednesday, October 27, 2021 at 2:41:29 AM UTC+13, Arne Vajhøj wrote:
> On 10/26/2021 2:17 AM, Lawrence D’Oliveiro wrote:
>
>> Whereas Microsoft still seem to believe that 26 drive letters ought to be enough
>> for anybody.
>
> The English language decided on 26 letters.

But the idea to only use one letter to identify a drive came from Gary Kildall with CP/M. Which, I understand, he cross-developed on a DEC PDP-10 running TOPS-10, which had multicharacter names for both disk and non-disk devices. (As did every subsequent DEC OS.)

Then Microsoft carried over the same limitation with MS-DOS for some reason, even though that was ostensibly running on “16-bit” hardware by now, not subject to the same memory limitations as the old “8-bit” machines.

> The last Windows server version that required a unique drive letter per
> disk was NT 4.0 Server (released 1996, end of support 2002/2004). All
> newer Windows versions (2003, 2008, 2008R2, 2012, 2012R2, 2016, 2019
> and 2022) supports volume mount points.

Which are too fiddly for ordinary people to use, can cause incompatibilities with some software, are tricky to set up for things like USB sticks or network shares, etc, etc.

[alanfe...@gmail.com]

On Tuesday, October 26, 2021 at 6:31:27 PM UTC-4, Lawrence D’Oliveiro wrote:
> On Tuesday, October 26, 2021 at 11:19:27 PM UTC+13, alanfe...@gmail.com wrote:
> > On Monday, October 25, 2021 at 10:48:26 PM UTC-4, Lawrence D’Oliveiro wrote:
> >> On Monday, October 25, 2021 at 5:28:47 PM UTC+13, alanfe...@gmail.com wrote:
> >>> When was EDT updated to allow more than 22 lines? IDK, but it's made me happy!
> >> “EDT” and “happy” were never words I considered using in the same sentence ...
> > I understand. Use the right tool for the job, I say. For me that is usually EDT.
> What drove me mad about EDT was its insistence on a “current direction” mode. Instead of having separate movement commands for going forward and backward by word, line etc, it had a single command for each, with separate “forward” and “backward” keys for setting the direction.
>
> But since I could never remember what was the last direction key I hit, I would always have to hit it again, so every movement command became *two* keystrokes instead of one.

Hmmm. Somehow this never bothered me much. I usually start with key 4 or 5 if I haven't touched the keypad in a short while. It's pretty quick, unless you're the hunt-and-peck type.

If you want to scroll up or down several pages, you only need to hit 4 or 5 once, then just repeat the 8 key until you find what you want.

But I fully understand why some _are_ bothered by it. Hey, typing directory brackets used to annoy me. And at least one other, who liked my TO.COM (an early rather primitive version) for that reason alone! Anyway, after a few years of typing long paths in Unix commands, I'm not bothered by it nearly as much. I actually tried to write a cd program in Unix, but Unix thwarted me at every turn. First, it would spawn a process to run it and change the directory therein, exit, and I'm still in the same directory! I was told a trick to take care of that, when something else screwed me over. So after a short while I punted and just typed out the long bloody paths. Arghh.

> > One thing I really like about EDT is the key definitions. You can see them and change
> > them without needing to recompile and make a new huge TPU section file ...

> Emacs does all that, and more. It also has the advantage of being built on LISP, which is still one of the most advanced programming languages around.

So Emacs lets you put in key definitions in a readable file and you don't need to compile it into a huge section file and thereafter not be able to list your commands? If it does, that's cool!

I've used emacs on occasion. A lot of two-keys-at-the-same-time commands. And often two of _them_ in a row. And the 4,5 bit bothers you on EDT? Here's an emacs key definition with three in a row!: C-x 4 C-o

OTOH, I _like_ EDT's command mode. You see the asterisk at the bottom or you don't. You can search for a term and just see the lines that have that term, like with the SEARCH command, instead of skipping around the file like you have to on PCs and Macs. Well, Excel does have Find All, except the Mac version doesn't! Arghh.

And another problem I always had with emacs: Once you start a session for the first time in a long time, you can't figure out how to exit. Well, I'd always forget it after a while and have to look it up. You'd think it would be a single key, like Ctrl-X or something, but NOOOOOOO. It's C-x C-c according to https://www.gnu.org/software/emacs/refcards/pdf/refcard.pdf Who could remember that after several months of not using emacs? What does the c stand for anyway?

Hey, if emacs works for you, that's great. Go for it.

AEF

[alanfe...@gmail.com]

I don't have any fields, but I _do_ have electric lights! Most of them are incandescent though. Only a very few LEDs. Hey, incandescents are better if heat goes down. Happened to me once!

Hey, if you need a blow torch instead of a match, go for it!!!

[Phillip Helbig (undress to reply)]

In article <4016c182-65dc-421c...@googlegroups.com>,
=?UTF-8?Q?Lawrence_D'Oliveiro?= <lawren...@gmail.com> writes:

> What drove me mad about EDT was its insistence on a \_current
> direction_/ mode.

It's not a bug, it's a feature. (By the way, your 8-bit characters have
been substituted with 7-bit replacements to be more usenet-friendly, via
an EDT macro.)

But you can have what you want! Say you want to SEARCH for a string:

PF1 PF3 <string> ENTER

is probably what you're doing. But

PF1 PF3 <string> KP4

is "find in forward direction"

and

PF1 PF3 <string> KP5

is "find in reverse direction.

ENTER is, as usual, "FIND NEXT" in the current direction.

So you can change (or confirm) the current direction at the same time as
you tell it to search, all with one keystroke. That's what I call
efficiency.

[Phillip Helbig (undress to reply)]

In article <f48435df-7ad5-4163...@googlegroups.com>,

"alanfe...@gmail.com" <alanfe...@gmail.com> writes:

> Hmmm. Somehow this never bothered me much. I usually start with key 4 or 5
> if I haven't touched the keypad in a short while. It's pretty quick, unless
> you're the hunt-and-peck type.

Indeed. I have saved literally years of my life because I learned to
touch-type when I was about 12. When I learned VMS, I extended that to
the keypad for EDT, MAIL, the debugger, and NEWSRDR, as well as keys I
have defined myself (consistently) in these and other applications.

Life is too short not to touch type.

[Phillip Helbig (undress to reply)]

In article <ba4e4aeb-b8c0-44b7...@googlegroups.com>,

"alanfe...@gmail.com" <alanfe...@gmail.com> writes:

> I don't have any fields, but I _do_ have electric lights!

As Maxwell said, I have an electric field.

[Lawrence D’Oliveiro]

On Wednesday, October 27, 2021 at 3:34:29 PM UTC+13, alanfe...@gmail.com wrote:
> So Emacs lets you put in key definitions in a readable file and you don't need to
> compile it into a huge section file and thereafter not be able to list your commands?
> If it does, that's cool!

It’s all just a bunch of ELISP definitions. Your custom functions and commands can have docstrings as well, to integrate with the built-in help system.

Here is my current collection of public customizations: <https://github.com/ldo/emacs-prefs>.

> I've used emacs on occasion. A lot of two-keys-at-the-same-time commands. And often
> two of _them_ in a row. And the 4,5 bit bothers you on EDT? Here's an emacs key
> definition with three in a row!: C-x 4 C-o

Less common commands take more keystrokes, more common ones (like the movement ones I mentioned) typically take fewer.

> And another problem I always had with emacs: Once you start a session for the first time
> in a long time, you can't figure out how to exit. Well, I'd always forget it after a while and
> have to look it up. You'd think it would be a single key, like Ctrl-X or something, but

> NOOOOOOO. It's C-x C-c ...

I tried for a while, binding it to a single custom keystroke. I removed that after I hit it by accident once too often.

The thing with Emacs is, just keep it running all the time. And also turn on its server mode. I have the following command defined in Bash:

    alias e='emacsclient -n'

So in any terminal session, I can just type “e «filename»” to have that file open immediately in the always-available Emacs window.

> Hey, if emacs works for you, that's great. Go for it.

There are depths to it I haven’t explored yet. For example, the text-attribute and overlay features (there seems to be some overlap between the two). One of the things I have done with overlays is being able to collapse (hide) selected regions of a file.

[alanfe...@gmail.com]

He had two -- zzzzt -- two --zzzzt -- two fields in one!

Some of you younger folk might not get the reference.

[Dave Hayter]

On Monday, 25 October 2021 at 08:59:32 UTC+1, Jan-Erik Söderholm wrote:
> Den 2021-10-25 kl. 06:28, skrev alanfe...@gmail.com:
> > On Sunday, October 24, 2021 at 12:22:44 PM UTC-4, Stephen Hoffman wrote:
> >> On 2021-10-22 06:45:02 +0000, alanfe...@gmail.com said:
> >>
> >>> This used to be broken. Strange things would happen under certain
> >>> conditions, like only the directory being changed or the directory
> >>> being hidden in sys$disk. It appears to be fixed on EISNER. Is there
> >>> documentation of this fix somewhere? I've been searching on the web,
> >>> only to come up empty. TIA! (^_^(
> >> *You rang?*
> >
> > Yep! That was the best part of the show, no? When Lurch would suddenly turn into the scene seemingly out of nowhere and say, "You rang?"
> > [...]
> >>
> >> Logical names are volatile key-value store design and straight out of
> >> ~1984, and long since and widely used and misused.
> >>
> >> For device redirection, they're adequate.
> >>
> >> Logical names have had "surprising" edge cases ~forever, too.
> >>
> >> The whole of the default directory stuff cited is just utterly
> >> hilarious, but that design is just too entrenched to ever be fixed and
> >> the existing stuff deprecated and removed.
> >
> > But Hoff, it _was_ fixed! Looks like no one knows when or what version of VMS it was fixed in.
> >
> > Here's a bonus question. When was EDT updated to allow more than 22 lines? IDK, but it's made me happy!
> There as an EDT patch just a year ago or so from VSI.
> Had something with numnber of lines to do, I think...
> >
> > Bonus No. 2: In what version of VMS was "Proactive Reclamation of Memory from Idle Processes" introduced? This one I found on my own. Turns out it was v5.4-3. Made a big difference in my main system at the time!
> >>
> > [...]
> >> --
> >> Pure Personal Opinion | HoffmanLabs LLC
> > AEF
> >
Damn! What a pity I missed out on the EDT update! (I retired 18 months ago)
EDT was my favourite editor by a very long way.

[David Jones]

On Wednesday, October 27, 2021 at 6:38:37 AM UTC-4, dhay...@gmail.com wrote:
> EDT was my favourite editor by a very long way.

30+ years ago, I wrote ~200 lines of TPU to customize the EVE global section to be
a blend of EVE and EDT interfaces. I had to tweak it in 1989 for the V5 TPU, but
since have used the exact same TPU$SECTION file ever since.

[Arne Vajhøj]

On 10/26/2021 6:46 PM, Lawrence D’Oliveiro wrote:
> On Wednesday, October 27, 2021 at 2:41:29 AM UTC+13, Arne Vajhøj wrote:
>> The last Windows server version that required a unique drive letter per
>> disk was NT 4.0 Server (released 1996, end of support 2002/2004). All
>> newer Windows versions (2003, 2008, 2008R2, 2012, 2012R2, 2016, 2019
>> and 2022) supports volume mount points.
>
> Which are too fiddly for ordinary people to use,

Likely.

But then I do not expect "ordinary people" to be responsible for
"entrust mission-critical business functions to" systems.

And if they are then I would not blame MS for the consequences.

> can cause incompatibilities with some software,

Invisible to software.

> are tricky to set up for things like USB sticks

It may be a problem to stuff 30 USB sticks in a Windows
server.

But believe me that is not a problem for "entrust mission-critical
business functions to" systems.

Arne

[Arne Vajhøj]

Very similar experience here.

Hacked the entire EVE code in 4.4 in 87. Rewrote to build
on top of EVE code in 5.5 in 89.

A bit more lines as I wanted it to be more CDC NOS FSE like.

Arne

[Simon Clubley]

Ok, at this point, does anyone want to confess to writing an Edlin
clone in TPU ? :-)

[Arne Vajhøj]

On 10/27/2021 9:44 AM, Arne Vajhøj wrote:
> Very similar experience here.
>
> Hacked the entire EVE code in 4.4 in 87. Rewrote to build
> on top of EVE code in 5.5 in 89.

5.0 of course - not 5.5.

Arne

[Lawrence D’Oliveiro]

On Thursday, October 28, 2021 at 2:30:00 AM UTC+13, Arne Vajhøj wrote:
>
> On 10/26/2021 6:46 PM, Lawrence D’Oliveiro wrote:
>>
>> On Wednesday, October 27, 2021 at 2:41:29 AM UTC+13, Arne Vajhøj wrote:
>>>
>>> The last Windows server version that required a unique drive letter per
>>> disk was NT 4.0 Server (released 1996, end of support 2002/2004). All
>>> newer Windows versions (2003, 2008, 2008R2, 2012, 2012R2, 2016, 2019
>>> and 2022) supports volume mount points.
>>
>> Which are too fiddly for ordinary people to use,
>
> Likely.
>
> But then I do not expect "ordinary people" to be responsible for
> "entrust mission-critical business functions to" systems.

How about a system where the hot-plug mount points are set up automatically as part of system defaults?

>> can cause incompatibilities with some software,
>
> Invisible to software.

You might think so. But then consider proprietary software installers that check for available space at the top level of the C: drive, and refuse to proceed if there’s no room there, even if they are actually going into another mount point.

>> are tricky to set up for things like USB sticks
>
> It may be a problem to stuff 30 USB sticks in a Windows
> server.

It may be a problem anything business-critical on a Windows server going forward, at least the non-cloud version, given that Microsoft is winding back its development efforts on that.

Another thing is that the mount-point feature is specific to NTFS. Unlike Linux, which has a generic VFS layer which handles things like that, for some reason Microsoft tied up important features like this with one particular filesystem, so it won’t work with anything else.

[alanfe...@gmail.com]

On Monday, October 25, 2021 at 9:25:10 PM UTC-4, alanfe...@gmail.com wrote:
> On Monday, October 25, 2021 at 3:47:23 PM UTC-4, Dave Froble wrote:
> > On 10/25/2021 2:28 PM, alanfe...@gmail.com wrote:
> >
> > > But it does a lot more than that. SET DEFAULT had 2 or 3 silly bugs that I figured wouldn't take long to fix.
> > What bugs? I've never seen any.
> I posted on earlier in the thread. I think it's even mentioned in the docs once, but I don't recall where and doubt I can find it.
>
> Here is my summary of them:
> The DCL program SET DEFAULT has two problems with nested logical
> names:
>
>
> 1.) If the first translation has a trailing colon and there is no
> explicit directory-spec in the second translation, SYS$DISK is changed
> to the 1st translation and the directory portion is not changed. The
> "actual" current directory remains hidden in SYS$DISK.
> [ I assume this mean if you specify <LNM>:[directory] ]
> 2.) If the 1st translation has no trailing colon, then only the
> directory portion of the default is changed! This can leave SYS$DISK
> incorrect thereby leaving the default incorrect.
> [This is the example posted by someone long ago which I put in one of my first posts in this thread.]
> Additionally, in certain circumstances, SET DEFAULT returns an error
> but still changes your process to a bad default.
> [Unfortunately, I don't recall any examples of this.]

> > --
> > David Froble Tel: 724-529-0450
> > Dave Froble Enterprises, Inc. E-Mail: da...@tsoft-inc.com
> > DFE Ultralights, Inc.
> > 170 Grimplin Road
> > Vanderbilt, PA 15486
> AEF

The main problem happens when your equivalence name is a logical name and you don't include the colon. For example:

$ DEFINE HOME SYS$LOGIN

is potentially troublesome. The better way is

$ DEFINE HOME SYS$LOGIN:

I found that if you always include the colon as above, you don't get into trouble.
Still, there was some rare case where you get a fatal error but it set your default anyway. Unfortunately, I don't recall what that was. If I stumble upon it later I will post it here.

[alanfe...@gmail.com]

On Monday, October 25, 2021 at 12:28:47 AM UTC-4, alanfe...@gmail.com wrote:
[...]

>
> Here's a bonus question. When was EDT updated to allow more than 22 lines? IDK, but it's made me happy!

Found it! It's

vms842l2a_edt-v0100.release_notes

The info about the SET DEFAULT bugs and word that it would be fixed in a later version of VMS is in

vmsu2055_release_notes

Still looking for the release notes where SET DEFAULT and SHOW DEFAULT were fixed.

>
[...]
> AEF

[Lawrence D’Oliveiro]

On Thursday, October 28, 2021 at 2:30:00 AM UTC+13, Arne Vajhøj wrote:
> But believe me that is not a problem for "entrust mission-critical
> business functions to" systems.

There are many mission-critical business functions running on Linux servers making use of storage technologies like SANs and LVM, which can lead to a proliferation of mounted volumes. Also remember that, on Linux, the visibility of mounted filesystems can be segregated into “filesystem namespaces”, such that a process can only see those filesystems within its assigned filesystem namespace.

This is a key part of container technologies, like Docker. Microsoft even tried to support Docker under Windows for a while, but it seems to have given up. Wonder why ...

[Arne Vajhøj]

On 10/31/2021 5:05 PM, Lawrence D’Oliveiro wrote:
> This is a key part of container technologies, like Docker. Microsoft
> even tried to support Docker under Windows for a while, but it seems
> to have given up. Wonder why ...

MS still support docker. On Windows Server 2016, 2019 and 2022.

Arne

[Lawrence D’Oliveiro]

For certain values of “support” <https://www.theregister.com/2019/06/18/the_future_of_docker_on_windows_is_linux_says_docker/>:

    At Microsoft's Build conference last month, Gabe Monroy, lead program
    manager for the Azure Container Compute team, was asked whether
    Windows Containers are for legacy and Linux Containers for new
    projects. "I think that is a fair description," he said ...

As for Windows Server 2022 <https://www.theregister.com/2021/07/28/windows_server_2022_sac/>:

    At a high level, it seems that Microsoft is investing in Azure and
    AKS, and that on-premises Windows Server users are in some respects
    being left behind. It is notable, for example, that Active Directory's
    functional level has not changed since Windows Server 2016, and this
    remains the case in Server 2022.

...

    SMB over QUIC, which promises secure access to Windows file shares
    from anywhere, is another key feature which will be available only in
    Windows Server Azure Edition.

[Arne Vajhøj]

On 10/31/2021 6:22 PM, Lawrence D’Oliveiro wrote:
> On Monday, November 1, 2021 at 10:34:07 AM UTC+13, Arne Vajhøj wrote:
>>
>> On 10/31/2021 5:05 PM, Lawrence D’Oliveiro wrote:
>>>
>>> This is a key part of container technologies, like Docker. Microsoft
>>> even tried to support Docker under Windows for a while, but it seems
>>> to have given up. Wonder why ...
>>
>> MS still support docker. On Windows Server 2016, 2019 and 2022.
>
> For certain values of “support” <https://www.theregister.com/2019/06/18/the_future_of_docker_on_windows_is_linux_says_docker/>:
>
>     At Microsoft's Build conference last month, Gabe Monroy, lead program
>     manager for the Azure Container Compute team, was asked whether
>     Windows Containers are for legacy and Linux Containers for new
>     projects. "I think that is a fair description," he said ...

Here is free mini course in IT terminology:
* supported means that the vendor accept bug reports and create fixes
* legacy means that it works but is not seen as a good choice
for new stuff

Being legacy does not imply not supported.

Arne

[Arne Vajhøj]

On 10/31/2021 6:22 PM, Lawrence D’Oliveiro wrote:
> As for Windows Server 2022 <https://www.theregister.com/2021/07/28/windows_server_2022_sac/>:
>
>     At a high level, it seems that Microsoft is investing in Azure and
>     AKS, and that on-premises Windows Server users are in some respects
>     being left behind. It is notable, for example, that Active Directory's
>     functional level has not changed since Windows Server 2016, and this
>     remains the case in Server 2022.
>
> ...
>
>     SMB over QUIC, which promises secure access to Windows file shares
>     from anywhere, is another key feature which will be available only in
>     Windows Server Azure Edition.

I suspect that both TheRegister writer and you did not really
understand the content of that article.

MS offers 3 different options:
* Azure software hosted by MS
* Azure software hosted on-premise
* traditional stuff hosted on-premise

Thew new features goes into the Azure software. But that
does not mean no support for on-premise. It just means that
on-premise will have to function cloud-like.

Arne

[Lawrence D’Oliveiro]

On Monday, November 1, 2021 at 12:15:55 PM UTC+13, Arne Vajhøj wrote:
> Here is free mini course in IT terminology:
> * supported means that the vendor accept bug reports and create fixes

When was the last time you got Microsoft to accept a bug report and issue a fix?

[Lawrence D’Oliveiro]

On Monday, November 1, 2021 at 12:24:29 PM UTC+13, Arne Vajhøj wrote:
> MS offers 3 different options:
> * Azure software hosted by MS
> * Azure software hosted on-premise
> * traditional stuff hosted on-premise
>
> Thew new features goes into the Azure software. But that
> does not mean no support for on-premise. It just means that
> on-premise will have to function cloud-like.

Those extra features are simply not available in the third option.

[Arne Vajhøj]

I have not.

But other people have.

Some people experience a problem and report it and MS fix it.

Some white hats find a vulnerability and send it in and MS fix it.

Some black hats find and exploit a vulnerability and MS fix it.

That is what MS does. What IBM does. What Redhat does. What Oracle
does. What various open source projects do.

Arne

[John Dallman]

In article <d889125f-9231-40dd...@googlegroups.com>,

lawren...@gmail.com (Lawrence D_Oliveiro) wrote:

> When was the last time you got Microsoft to accept a bug report and
> issue a fix?

Last year. I reported several optimiser bugs in their C/C++ compiler for
ARM64. They fixed some and supplied workarounds for the others. I have
two advantages in doing this:

I work for a large company (I can speak more freely if I don't say who),
which pays a lot for Microsoft support. That gets you listened to, as
does being polite and clear. We don't get charged for fixes that are due
to Microsoft bugs, and they are pretty reasonable about making that
decision.

I have been reporting compiler bugs to Microsoft since ... checks ...
1997, and have a decent working relationship with them. My reports are
clear, and usually correct, and they act on them quite promptly.

John

[Lawrence D’Oliveiro]

On Tuesday, November 2, 2021 at 4:31:53 AM UTC+13, Arne Vajhøj wrote:
>
> On 11/1/2021 1:55 AM, Lawrence D’Oliveiro wrote:
>>
>> On Monday, November 1, 2021 at 12:15:55 PM UTC+13, Arne Vajhøj wrote:
>>>
>>> Here is free mini course in IT terminology:
>>> * supported means that the vendor accept bug reports and create fixes
>>
>> When was the last time you got Microsoft to accept a bug report and issue a fix?
>
> I have not.
>
> But other people have.

Any of them with Docker for Windows?

Because if not, then by your own definition of “support”, you have not demonstrated that Microsoft is offering “support” for Docker for Windows.

[Lawrence D’Oliveiro]

On Tuesday, November 2, 2021 at 9:20:54 AM UTC+13, John Dallman wrote:
> I work for a large company ... which pays a lot for Microsoft support.

Is that on top of the actual licence fees for Windows itself?

Just to make clear what ordinary folks are (not) getting for the money they pay to Microsoft ...

[John Dallman]

In article <c9d02312-bf2a-4843...@googlegroups.com>,

lawren...@gmail.com (Lawrence D_Oliveiro) wrote:
> On Tuesday, November 2, 2021 at 9:20:54 AM UTC+13, John Dallman
> wrote:
> > I work for a large company ... which pays a lot for Microsoft
> > support.
> Is that on top of the actual licence fees for Windows itself?

Yes.

I suspect if someone who didn't have that kind of support were to report
compiler bugs clearly, the bugs would get attention, although not
hotfixes, which are one of the things you get for the support fees.

However, few people know how to report compiler bugs effectively, and
I've seen internal examples within companies that are hilariously bad.
Handing over a 50KB .c file generated by a transpiler from another
language with details of "I think it goes wrong somewhere in here" is not
the way to get things fixed.

John

[alanfe...@gmail.com]

On Wednesday, October 27, 2021 at 12:32:09 AM UTC-4, Phillip Helbig (undress to reply) wrote:
> In article <4016c182-65dc-421c...@googlegroups.com>,
> =?UTF-8?Q?Lawrence_D'Oliveiro?= <lawren...@gmail.com> writes:
>
> > What drove me mad about EDT was its insistence on a \_current
> > direction_/ mode.
>
> It's not a bug, it's a feature. (By the way, your 8-bit characters have
> been substituted with 7-bit replacements to be more usenet-friendly, via
> an EDT macro.)
>
> But you can have what you want! Say you want to SEARCH for a string:
>
> PF1 PF3 <string> ENTER
>
> is probably what you're doing. But
>
> PF1 PF3 <string> KP4
>
> is "find in forward direction"
>
> and
>
> PF1 PF3 <string> KP5
>
> is "find in reverse direction.
>
> ENTER is, as usual, "FIND NEXT" in the current direction.

Why Press Enter? Why not press PF3. That way your finger is already in place to find the next one!

>
> So you can change (or confirm) the current direction at the same time as
> you tell it to search, all with one keystroke. That's what I call
> efficiency.

AEF

[alanfe...@gmail.com]

On Monday, October 25, 2021 at 3:47:23 PM UTC-4, Dave Froble wrote:
> On 10/25/2021 2:28 PM, alanfe...@gmail.com wrote:
>
> > But it does a lot more than that. SET DEFAULT had 2 or 3 silly bugs that I figured wouldn't take long to fix.


> What bugs? I've never seen any.

Here are some SET DEFAULT bugs:

From vmsu2055_release_notes.txt

1.1.5 SET DEFAULT and SHOW DEFAULT Commands-Behavior Changed

V5.5-2

In VMS Version 5.5-2, the DCL command SET DEFAULT does not
set the default correctly when you use a logical name that
has multiple translations. For example:

$ SET DEFAULT DISK1:[DIRECTORY1]
$ DEFINE myfile myfile1
$ DEFINE myfile1 DISK2:[DIRECTORY2]
$ SET DEFAULT myfile
$ SHOW DEFAULT
DISK1:[DIRECTORY2]

In this example, the default should read:

DISK2:[DIRECTORY2]

This problem will be corrected in a future release of VMS.

Currently, DCL does not show the right default device in
the SHOW DEFAULT command when you use a logical name that
has multiple translations. For example:

$ SET DEFAULT DISK1:[DIRECTORY]
$ DEFINE myfile myfile1
$ DEFINE myfile1 DISK2:[DIRECTORY]
$ SET DEFAULT myfile: ! Must be with a colon
$ SHOW DEFAULT
myfile:[DIRECTORY]
$ WRITE SYS$OUTPUT F$ENVIRONMENT("DEFAULT")
myfile:[DIRECTORY]

In this example, "myfile" is used as the default
device. The correct default directory display should be
DISK2:[DIRECTORY].

This SHOW DEFAULT behavior will be changed in a future
release of VMS.
======================================================================
The bug in the first example has been fixed. I don't know when except
that it was probably 7.0 or later.

The second one is still broken on OpenVMS V8.4-2L2.

Regardless, you won't encounter these problems if you always use
TO.COM to change your default. (TO.COM v5.3 will be released soon!)
======================================================================
======================================================================
OLD:

The DCL program SET DEFAULT has two problems with nested logical
names:

1.) If the first translation has a trailing colon and there is no
explicit directory-spec in the second translation, SYS$DISK is changed
to the 1st translation and the directory portion is not changed. The
"actual" current directory remains hidden in SYS$DISK.

The above is still broken as of OpenVMS V8.4-2L2.

Got new details on this. SET DEF keeps translating until it gets to a
logical name with a trailing colon. Then SYS$DISK is changed to that
logical name, leaving the directory unchanged; unless it contains a
device and directory, in which case you get just that: the device and
the directory upon running SHOW DEFAULT. Note that if you include a
trailing colon in the SET DEFAULT command, the logical name will be
assigned to SYS$DISK, just as if it were a search list.

Example:

$ SHOW DEF
DSA3:[DECUSERVE_USER.FELDMAN.TO-V530]
$ SHOW LOG A
"A" = "B:" (LNM$PROCESS_TABLE)
1 "B" = "C:" (LNM$PROCESS_TABLE)
2 "C" = "D:" (LNM$PROCESS_TABLE)
3 "D" = "SYS$LOGIN:" (LNM$PROCESS_TABLE)
4 "SYS$LOGIN" = "DISK_USER:[DECUSERVE_USER.FELDMAN]" (LNM$JOB_831D6780)
$ SET DEF A
$ SHOW DEF
B:[DECUSERVE_USER.FELDMAN.TO-V530]
$ SET DEF D:
$ SH DEF
D:[DECUSERVE_USER.FELDMAN.TO-V530]
$ SET DEF D
$ SH DEF
DISK_USER:[DECUSERVE_USER.FELDMAN]
$

2.) If the 1st translation has no trailing colon, then only the
directory portion of the default is changed! This can leave SYS$DISK
incorrect thereby leaving the default incorrect.

The above, no 2., has been fixed. I don't know when, though.

The above, no 2., has been fixed. I don't know when, though.

Additionally, in certain circumstances, SET DEFAULT returns an error

but still changes your process to a bad default. But I haven't been
able to reproduce this on EISNER (OpenVMS V8.4-2L2). Maybe it's been
fixed.

UPDATE on 2022/10/16

It's still broken:

$ SH DEF
DISK_USER:[DECUSERVE_USER.FELDMAN]
$ SET DEF DSA2:DSA3
%RMS-F-DIR, error in directory name
$ SH DEF
DSA2:[DECUSERVE_USER.FELDMAN]
%DCL-I-INVDEF, DSA2:[DECUSERVE_USER.FELDMAN] does not exist
$

As you can see, SET DEFAULT produced a fatal error, yet still changed
the default. Yeah, it's unlikely you'll ever type something like that.
Maybe there's another type of typo that causes this.

Anyway,

TO.COM fixes all of the above problems and does much more. See the
other .TXT files for additional information.

TO.COM v5.3 to be released soon!

> --
> David Froble Tel: 724-529-0450
> Dave Froble Enterprises, Inc. E-Mail: da...@tsoft-inc.com
> DFE Ultralights, Inc.
> 170 Grimplin Road
> Vanderbilt, PA 15486

Alan Feldman