Re: pty behaviour

Marcus Brinkmann <[email protected]> Sun, 07 Mar 2004 00:29:16 +0100
Newsgroups gmane.os.hurd.devel.readers
Message-ID <[email protected]>
At 06 Mar 2004 15:10:16 -0800,
Thomas Bushnell, BSG wrote:
> 
> Roland McGrath <[email protected]> writes:
> 
> > It's possible that Linux is behaving the same way and just keeping up
> > better (and we already know it's much faster).  (When I am in the mood, I
> > can read Linux source and see what exactly it does.)  That is, the 8kish
> > value is how much cumulatively gets through over many iterations of the
> > consumer almost but not quite keeping up with the producer, before the
> > small buffer of the remainders adds up to being full.
> 
> Possibly, but I would expect the "paste" operation to simply generate
> one big write request.

(Uhm, didn't get Roland's mail yet).  The perl IO::Tty test suite
specifically checks how much can be written at once, and it gave me
8061 as the limit.

Thanks,
Marcus