info
Google Groups no longer supports new Usenet posts or subscriptions. Historical content remains viewable.
Dismiss
Scrambled Threads
3 views
Skip to first unread message
croy
Jun 18, 2021, 12:56:44 PM6/18/21
to
Recently, Agent seems to be scrambling threads: a recent message in a decades-old thread, or a
decades old message in a recent thread. The subject is very different, but I usually don't
notice that until I've spents a minute or two trying to screw my brain back together.
--
croy
decades old message in a recent thread. The subject is very different, but I usually don't
notice that until I've spents a minute or two trying to screw my brain back together.
--
croy
Geoff
Jun 18, 2021, 3:52:34 PM6/18/21
to
On Fri, 18 Jun 2021 09:56:37 -0700, croy <cr...@spam.invalid.net>
wrote:
>Recently, Agent seems to be scrambling threads: a recent message in a decades-old thread, or a
>decades old message in a recent thread. The subject is very different, but I usually don't
>notice that until I've spents a minute or two trying to screw my brain back together.
Did you accidentally turn off threaded sort? The little + icon to the
right of the Status column in the message list pane should be
depressed for threaded sort.
wrote:
>Recently, Agent seems to be scrambling threads: a recent message in a decades-old thread, or a
>decades old message in a recent thread. The subject is very different, but I usually don't
>notice that until I've spents a minute or two trying to screw my brain back together.
right of the Status column in the message list pane should be
depressed for threaded sort.
Arthur T.
Jun 18, 2021, 5:39:51 PM6/18/21
to
In Message-ID:<itjpcg5k9fb6s3tac...@4ax.com>,
Do you have a lot of messages in that newsgroup? If so, this
conversation from long ago might be applicable:
>>This is an Agent bug.
>>Welcome to the wonderful world of "hash collisions".
>It's a great improvement over the world of "what the heck?".
If that's the problem, the solution is reduce the number of messages
you've kept in that group. When Usenet was more active, I used to
create annual folders to hold the old messages I wanted to keep.
--
Arthur T. - ar23hur "at" pobox "dot" com
conversation from long ago might be applicable:
>>This is an Agent bug.
>>Welcome to the wonderful world of "hash collisions".
>It's a great improvement over the world of "what the heck?".
If that's the problem, the solution is reduce the number of messages
you've kept in that group. When Usenet was more active, I used to
create annual folders to hold the old messages I wanted to keep.
--
Arthur T. - ar23hur "at" pobox "dot" com
croy
Jun 18, 2021, 11:39:32 PM6/18/21
to
--
croy
croy
Jun 18, 2021, 11:42:23 PM6/18/21
to
On Fri, 18 Jun 2021 17:39:51 -0400, Arthur T. <art...@munged.invalid> wrote:
>In Message-ID:<itjpcg5k9fb6s3tac...@4ax.com>,
>croy <cr...@spam.invalid.net> wrote:
>
>>Recently, Agent seems to be scrambling threads: a recent message in a decades-old thread, or a
>>decades old message in a recent thread. The subject is very different, but I usually don't
>>notice that until I've spents a minute or two trying to screw my brain back together.
>
>Do you have a lot of messages in that newsgroup? If so, this
>conversation from long ago might be applicable:
>
>>>This is an Agent bug.
>>>Welcome to the wonderful world of "hash collisions".
>>It's a great improvement over the world of "what the heck?".
234,259 in one group.
>In Message-ID:<itjpcg5k9fb6s3tac...@4ax.com>,
>croy <cr...@spam.invalid.net> wrote:
>
>>Recently, Agent seems to be scrambling threads: a recent message in a decades-old thread, or a
>>decades old message in a recent thread. The subject is very different, but I usually don't
>>notice that until I've spents a minute or two trying to screw my brain back together.
>
>Do you have a lot of messages in that newsgroup? If so, this
>conversation from long ago might be applicable:
>
>>>This is an Agent bug.
>>>Welcome to the wonderful world of "hash collisions".
>>It's a great improvement over the world of "what the heck?".
>If that's the problem, the solution is reduce the number of messages
>you've kept in that group. When Usenet was more active, I used to
>create annual folders to hold the old messages I wanted to keep.
--
croy
Arthur T.
Jun 19, 2021, 12:53:05 AM6/19/21
to
In Message-ID:<6ppqcglo5k48966tk...@4ax.com>,
hashes, but I've had (very rare) collisions at a tenth that size.
croy <cr...@spam.invalid.net> wrote:
>>Do you have a lot of messages in that newsgroup? If so, this
>>conversation from long ago might be applicable:
>>
>>>>This is an Agent bug.
>>>>Welcome to the wonderful world of "hash collisions".
>>>It's a great improvement over the world of "what the heck?".
>
>234,259 in one group.
A quick search doesn't show what size Agent uses for its message
>>Do you have a lot of messages in that newsgroup? If so, this
>>conversation from long ago might be applicable:
>>
>>>>This is an Agent bug.
>>>>Welcome to the wonderful world of "hash collisions".
>>>It's a great improvement over the world of "what the heck?".
>
>234,259 in one group.
hashes, but I've had (very rare) collisions at a tenth that size.
Arthur T.
Jun 19, 2021, 12:53:14 AM6/19/21
to
In Message-ID:<dmpqcg99f7c97f3ro...@4ax.com>,
Sorry, it was just too good a straight line to pass up.
croy <cr...@spam.invalid.net> wrote:
>>Did you accidentally turn off threaded sort? The little + icon to the
>>right of the Status column in the message list pane should be
>>depressed for threaded sort.
>
>
>It is still depressed.
Try singing it a song or buying it some ice cream.
>>Did you accidentally turn off threaded sort? The little + icon to the
>>right of the Status column in the message list pane should be
>>depressed for threaded sort.
>
>
>It is still depressed.
Sorry, it was just too good a straight line to pass up.
Ralph Fox
Jun 19, 2021, 1:44:44 AM6/19/21
to
On Fri, 18 Jun 2021 20:42:16 -0700, croy wrote:
> On Fri, 18 Jun 2021 17:39:51 -0400, Arthur T. <art...@munged.invalid> wrote:
>
>> In Message-ID:<itjpcg5k9fb6s3tac...@4ax.com>,
>> croy <cr...@spam.invalid.net> wrote:
>>
>>> Recently, Agent seems to be scrambling threads: a recent message in a decades-old thread, or a
>>> decades old message in a recent thread. The subject is very different, but I usually don't
>>> notice that until I've spents a minute or two trying to screw my brain back together.
>>
>> Do you have a lot of messages in that newsgroup? If so, this
>> conversation from long ago might be applicable:
>>
>>>> This is an Agent bug.
>>>> Welcome to the wonderful world of "hash collisions".
>>> It's a great improvement over the world of "what the heck?".
>
> 234,259 in one group.
This message from 1997 has a table showing the probability of a "hash
> On Fri, 18 Jun 2021 17:39:51 -0400, Arthur T. <art...@munged.invalid> wrote:
>
>> In Message-ID:<itjpcg5k9fb6s3tac...@4ax.com>,
>> croy <cr...@spam.invalid.net> wrote:
>>
>>> Recently, Agent seems to be scrambling threads: a recent message in a decades-old thread, or a
>>> decades old message in a recent thread. The subject is very different, but I usually don't
>>> notice that until I've spents a minute or two trying to screw my brain back together.
>>
>> Do you have a lot of messages in that newsgroup? If so, this
>> conversation from long ago might be applicable:
>>
>>>> This is an Agent bug.
>>>> Welcome to the wonderful world of "hash collisions".
>>> It's a great improvement over the world of "what the heck?".
>
> 234,259 in one group.
collision" in a group, vs. the number of headers in the group.
<https://groups.google.com/g/alt.usenet.offline-reader.forte-agent/c/zzElLwjhaS8/m/nB9Dv1dWJCwJ>
Google Groups mangles the layout spacing, so here is the original.
table. This is best viewed in a fixed-pitch font.
| >| 0.0001 0.001 0.01 0.1 0.5@ 0.9 0.99 0.999 0.9999
| >|-----------------------------------------------------------------------------------
| >| 67108786: 117 367 1162 3761 9646@ 17580 24861 30448 35157
>> If that's the problem, the solution is reduce the number of messages
>> you've kept in that group. When Usenet was more active, I used to
>> create annual folders to hold the old messages I wanted to keep.
>
> Hmmm. Thanks for that. I've never used folders that way in Agent.
--
Ralph
🦊
Ralph Fox
Jun 19, 2021, 2:07:35 AM6/19/21
to
On Sat, 19 Jun 2021 00:53:04 -0400, Arthur T. wrote:
> A quick search doesn't show what size Agent uses for its message
> hashes
See this 1997 message from Sherlog:
> A quick search doesn't show what size Agent uses for its message
> hashes
<https://groups.google.com/g/alt.usenet.offline-reader.forte-agent/c/VzjDcsZ-7b8/m/Yp9_H6T7du0J>
Message-ID: <3370f7be...@J.E.H.A.D>
For message-ids the hash is effectively 25 bits, not 26, in spite
of what that message says. The last character in the hash of a
message-id is always '>' and, because of the way the hash is
actually calculated, this effectively reduces the range from 26
bits to 25 bits for message-ids.
If "Enable threading by subject" is checked (the default) then one
can also get subject line hash collisions.
--
Kind regards
Ralph
🦊
Arthur T.
Jun 19, 2021, 3:07:39 AM6/19/21
to
In Message-ID:<u82rcgtinu8dc3eka...@4ax.com>,
The table you reproduced earlier seems to have been calculated based
on 26-bit hashes. Recomputed for 25-bit, I get these numbers for how
many messages needed for what chances of at least one hash collision
(in parens are the numbers I computed based on 26-bit hashes, which
match that other table):
50% 6821 (9646)
90% 12431 (17580)
99% 17579 (24861)
I didn't bother to redo the whole table, just these 3 points.
Ralph Fox <-rf-nz-@-.invalid> wrote:
>For message-ids the hash is effectively 25 bits, not 26, in spite
>of what that message says. The last character in the hash of a
>message-id is always '>' and, because of the way the hash is
>actually calculated, this effectively reduces the range from 26
>bits to 25 bits for message-ids.
Thank you for that information.
>For message-ids the hash is effectively 25 bits, not 26, in spite
>of what that message says. The last character in the hash of a
>message-id is always '>' and, because of the way the hash is
>actually calculated, this effectively reduces the range from 26
>bits to 25 bits for message-ids.
The table you reproduced earlier seems to have been calculated based
on 26-bit hashes. Recomputed for 25-bit, I get these numbers for how
many messages needed for what chances of at least one hash collision
(in parens are the numbers I computed based on 26-bit hashes, which
match that other table):
50% 6821 (9646)
90% 12431 (17580)
99% 17579 (24861)
I didn't bother to redo the whole table, just these 3 points.
0 new messages