-----You received this message because you are subscribed to the Google Groups "zfs-macos" group.To unsubscribe from this group and stop receiving emails from it, send an email to zfs-macos+...@googlegroups.com.For more options, visit https://groups.google.com/groups/opt_out.
Am Mittwoch, 3. April 2013 um 14:50 schrieb Lucien Pullen:
For consistency with Apple's products that use installers (namelyiTunes[1]), I think that full uninstall should be a terminal thing.We've already got fairly decent instructions that can be found by asearch engine easily.
I think this is a bit dangerous. With something as critical as afilesystem, updating shouldn't be something as casual as updating yourtext editor. ZFS /should/ be trivial to update, but only when the useris ready, not because the version checker is nagging them.
Something I would like to know about this is the update policy. How doyou ensure to the user that the release being nagged about is trulystable enough? Is this a scanner for whatever is "Featured"? or is it ablessed version?
When the port is ready, I recall you saying we were going to abandon theupstream numbers (or maybe I'm misremembering). As a one-time thing,how will the update check cope with a different numbering scheme? Notas important, but how do we convince users that "Y is more recent thanX" even if the number is lower?
How would you tune what day and time to run the scrub on? How would youcope with different pools that have different needs? Can you set afilesystem tree to not be included in (___) snapshots through theinterface? or is that something for a home-grown cron job?
Why just some of the options? If just some options, how do we determinewhat those are?
(x) ... your ideas here ...Not so much an idea as more of a possible issue.Is the preference pane written for the administrator? the user of asystem? or both?
>>> I think this is a bit dangerous. With something as critical as a
>>> filesystem, updating shouldn't be something as casual as updating
> your
>>> text editor. ZFS /should/ be trivial to update, but only when the user
>>> is ready, not because the version checker is nagging them.
That's precisely how casual and easy any stable release must be, or else it wouldn't be a stable release. It just periodically checks for the latest stable release via a simple URL and/or XML metadata, which in turn was issued to the web site against our existing, coherent release policy. Just like every other updating app in the world. By default, it would show the 'zpool status' data and could start a scrub, and it would have an 'advanced' button which is protected by the standard 'lock' icon, like any other preferences pane has. But, there is no magical development release listed in a default control panel nor any nagging; no free software does that, nor would there ever be any reason and I don't know where anyone got that idea. Your concerns are far too drastic, my friend!
>>> Something I would like to know about this is the update policy. How do
>>> you ensure to the user that the release being nagged about is truly
>>> stable enough? Is this a scanner for whatever is "Featured"?
> or is it a
>>> blessed version?
>> This is IMHO a philosophical question. But I think the normal user should
> see the stable releases, while the tester/thrill-seeker should be able to update
> to less well-tested versions, together with a big red warning sign of course.
There's nothing philosophical to it, just basic logic like every other piece of software. It's just stable (release) or unstable (prerelease/development). It's just metadata like I said.
>>> upstream numbers (or maybe I'm misremembering). As a one-time
> thing,
>>> how will the update check cope with a different numbering scheme? Not
>>> as important, but how do we convince users that "Y is more recent
> than
>>> X" even if the number is lower?
Coherent metadata would be issued against our existing, coherent release policy, as an every-time check. There is no guessing or worrying or marketing or brainwashing or ..... whatever the rest of this is all about. ;)
> script but rather write a background deamon that sleeps till it is ready to kick
> off whatever needs to be done.
Yes, to my limited knowledge, this is the silly and unreliable kludge that launchd requires of every arbitrary execution schedule. And then you hope that process didn't crash, that the system never ran out of memory and killed it, or that anything else unexpectedly happened. If I understand it correctly, I don't know how Apple thought this could possibly replace even 'cron'. I hope I'm mistaken.
See ya on the maczfs-devel list, hopefully.
This is the URL to the maczfs-devel list again, just in case that helps to find it. The easiest way I find it, is I surf to http://maczfs.org/ and click it on the bottom left.
Let's please take this discussion to the maczfs-devel list!