[NNTP] 64-bit article counter extension strawman

[Russ Allbery]

Steve Walker <nntp at nntpserver.com> writes:

> I didn't mean to step on your toe's.  I apologize if you saw it that
> way, it was not my intention.

Thank you... that was graciously said.  No troubles.

> The question (or I guess it was more of a statement) was are there any
> modern systems out there that can not use either the glibc or windows
> libraries?

Yes -- all versions of Unix other than Linux do not use glibc.  glibc is
not portable to versions of Unix other than Linux; some people have done
occasional one-off ports, but the maintainers of glibc disavow all
knowledge of them and refuse to support them in any way.  And extracting
and using portions of glibc is prohibitively difficult.

Trying to port old code and introduce things like 64-bit types is
something that unfortunately I have some experience with.  It's still very
obnoxious to do in a sane fashion.  If one just assumes Linux, or maybe
Linux and *BSD, it's much easier, but most Unix software maintainers are
loathe to do that.

> It work in the sense that everyone used a binary number that would hold
> a number large enough to get the job done.  Why didn't people use 8 or
> 16 bits numbers?  They choose a number that matched the current needs.
> If we left it open the same would happen again.  People would slowly
> migrate to the next larger binary number, in this case 63/64 bit.

The part of this that bothers me is that a server has no way of knowing
whether incrementing that article number just one more is going to cause a
bunch of clients to break (and possibly not just for that group).  But
perhaps I'm too fixated on this, and it's not really as much of an issue
as it feels like to me.

-- 
Russ Allbery (rra at stanford.edu)             <http://www.eyrie.org/~eagle/>