[5689] in www-talk@info.cern.ch
Re: holding connections open: a modest proposal
daemon@ATHENA.MIT.EDU (Darren New)
Mon Sep 19 00:13:14 1994
Date: Mon, 19 Sep 1994 06:05:58 +0200
Errors-To: listmaster@www0.cern.ch
Errors-To: listmaster@www0.cern.ch
Reply-To: dnew@sgf.fv.com
From: Darren New <dnew@sgf.fv.com>
To: Multiple recipients of list <www-talk@www0.cern.ch>
> There is another way. Some who code this other way have
> become fiends about it. Whether the "file" is the output of a program
> or not, the server could be coded as a pipeline. Not everyone wants
> to suck the whole thing into memory.
No. But not everyone wants to break it up, either.
By making the length bytes too short, you're preventing people from using
long lengths. By making the length bytes long enough, I'm not preventing
you from using short lengths.
> Not all OS's have file-to-memory mapping. Even when it's
> available it might not always give you the win you're after. If it's
> there, and you want to use it, and the situation allows it, great!
But the suggestion that everything be broken into 64K blocks prevents
that from working.
> But when suggesting protocol changes, consider that not everyone is
> going to use the same methods.
Nothing I suggested would prevent you from using any methods you
suggested. Your suggestions, however, would prevent me from using the
methods I suggested. That's all I'm saying. Just because you have a
4-byte length field doesn't mean you have to fill it up.
'Nuff said. This is getting way too silly, and I'm pretty sure I've taken
it to private email several times, only to have the replies resurface here.
Coulda been someone else, tho.
