Re: draft-cameron-tmux-02.txt

[Pete Cameron <[e-mail]>]

---------  Received message begins Here  ---------

> From: "Donald E. Eastlake 3rd (Beast)" <[e-mail]>
> 
> 
> >> Even if this caused degradation for a single ftp session, would it have
> >> increased throughput if there had been 10 ftp sesssions?
> >
> >It actually improved performance for 2 or more sessions to
> >and from a pair of hosts.
> 
> Is this mentioned in the draft anywhere?

No it isn't.

> If someone was running a heavily loaded ftp server with
> some clustering of clients, they might want to tmux to
> increase total service.

Yes that is true.  I think the following extract from an
email from Dave Crocker (from a message that was only sent
out to the authors of the draft) sums this up well (I hope
you don't mind Dave):

        just to underscore the 'simplicity' perspective that
        Pete cited: During the last IETF meeting, a number
        of interesting enhancements were discussed.  It was
        observed by one of the participants (can't remember
        who) that this facility is intended for a certain
        category of situation and really isn't intended to
        be highly general.  That doesn't mean that it should
        be designed in a crippled fashion, but rather that
        we should keep the goal in mind, when deciding how
        to deal with alternatives that are sticky.  (This
        reduces, of course, to a general philosophy of: if
        it doesn't cost anything, then add it; otherwise
        don't.)

Using TMux for FTP trafic comes under the "interesting
enhancements" category, so we made the decision to exclude
it. Making FTP work using TMux would cost something. If
someone wants to use TMux in this environment, it should
help, but that is not our goal.

Pete