[Top][All Lists]

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: [Monotone-devel] Question on layering

From: Nathaniel Smith
Subject: Re: [Monotone-devel] Question on layering
Date: Wed, 21 Feb 2007 21:02:22 -0800
User-agent: Mutt/1.5.13 (2006-08-11)

On Wed, Feb 21, 2007 at 04:25:09PM -0500, Ethan Blanton wrote:
> Now, I don't disagree that partial pull should be non-default behavior
> (necessarily), but I do disagree with the thesis that in most cases
> you don't need it.  This is very project-specific.

Yes.  I claim that most projects are the ones where you specifically
don't need it :-).  (I could be wrong, it's a perfectly empirical

> In the case of a
> Gaim import that I have been playing with, a pull of just the "trunk"
> branch of Gaim requires the network transfer of about 2.5x the size of a
> checked-out workspace, and three quarters of an hour or so elapsed
> wall clock time on a fast machine with a good network connection.
> This is a huge cost for a casual contributor who wants to produce a
> quick patch for an identified bug.
> Monotone is tiny, and even it takes forever to pull.  ;-)

The _time_ taken is simply a bug.  Let's be careful not to bake in
major design decisions as workarounds for an ephemeral (though
admittedly long-lived...) bug.

Network traffic is harder to fix.  I can't tell either way whether
that's going to be a problem long term... we might be able to optimize
for space more (there's at least some win possible just by using
stream compression instead of the current packet-by-packet
compression), and average bandwidth keeps getting wider...  Certainly
there will exist cases where the bandwidth is a problem, but the
question is which turns out to be the _common_ case, out of all
combinations of histories and network connetions...

-- Nathaniel

"The that sets have a very limited range of
activities -- they can't carry pianos, for example, nor drink

reply via email to

[Prev in Thread] Current Thread [Next in Thread]