ietf-822
[Top] [All Lists]

Re: mail vs. news ???

2003-02-23 16:56:39

"Bruce" == Bruce Lilly <blilly(_at_)erols(_dot_)com> writes:

 [message-ids]

 Bruce> If we're going to set a limit, we have to pick a number.
 Bruce> Let me try again, from a different perspective:

 Bruce> 1. The NNTP limit appears to be 501 octets, minus whetever the
 Bruce> longest command is, minus one

no, 510 minus the longest relevent command minus one, giving 501 as
the absolute maximum limit (due to TAKETHIS). The draft NNTP revisions
seem to have a limit of 12 characters for command names, which would
make the limit 497. In fact there's also an issue if you go over about
452 characters, when the ARTICLE or STAT response strings start to go
over 512 total; plus (I'd have to check the latest draft) possible
commands like "HDR someheader <id>".

Basically if you go over 501 then the transport collapses in ruin
(well, almost); whereas the 450-500 range just causes some minor
difficulty.

 Bruce> 3. So what's so special about 250, rather than 255 or 256?

it happens to be a value historically used by INN (present in various
1.x versions, then removed in 2.x, then explicitly added back when it
was discovered that yes, there really does have to be a limit
otherwise things break) and believed to be acceptable to other
software, while still giving plenty of headroom (message-ids of length
100-120 show up occasionally, ids of length 120-150 much more rarely,
that's about the longest seen in practice other than due to breakage
of some sort such as gateway loops).

In the absence of a really good reason (i.e. not merely aesthetic) to
use some other value, that's the one to pick.

-- 
Andrew.

<Prev in Thread] Current Thread [Next in Thread>