Re: INFO - Why

Brian Godette (bgodette@idcomm.com)
Wed, 03 Jun 1998 10:53:16 -0600


At 08:31 AM 6/2/98 -0400, you wrote:
>I am using ver 1.....I have notice the last few days that I could be on
>for ten minutes and then I will see again what one person had said all
>over again. Last night I saw the same thing four times from the same
>person. It usually happens with one person.
>
>Is that a bug in Ver 1...or have they setup something with wrong with
>their Ver. of CU...?
>

Yes and no, depends on how you look at it. So far the conditions I've seen
this happen under is thus. A reflector with a significant amount of chat
with only a few (often only one) people as the focus of the chat (think of
something like a press announcement Q&A and you get the idea). This results
in 1) alot of chat line IDs that *your* client has to keep track of, and 2)
one (or more) other clients with large recovery buffers. So my guess as to
what happens (unverified, but it does seem to be the case) is the internal
list of IDs keept by your client isn't large enough to hold the total
number of IDs seen over the last 60 seconds (the usual length of time the
respective flavors of CU hold chat for recovery), which results in your
client requesting recovery of chat lines it's actually seen before but
doesn't know about because the list isn't large enough. Clear as mud eh? :)

In anycase I have plans to deal with this problem with ERef, another "fix
the client" patch (only this time it seems to effect almost any flavor of
CU). As an added note, why you haven't seen this happen before is that
until Cornell 1.0 and WP 3.*, chat recovery was busted in CU if you were on
a PC (and not using 0.84b7) and weren't connected to an ERef.