RE: transfer-encoding

> -----Original Message-----
> From: Henrik Nordstrom [mailto:hno@hem.passagen.se]
> Sent: Friday, 5 January 2001 6:03 PM
> To: Robert Collins
> Cc: squid-dev@squid-cache.org
> Subject: Re: transfer-encoding
>
>
> Robert Collins wrote:
>
> > Yes I do - here's my reasoning...
> > I've found while working on the auth_rewrite stuff, that large scale
> > changes have a tendency to break dependant modules temporarily. I.E.
> > adding in a require module function. Personally, while I'm
> testing a new
> > idea out, I don't want to be breaking everything in sight until it's
> > stabilised - even if that's only a few hours work.
>
> This I fully agree on. However, once that testing is done the changes
> should be merged up to the base branch, and in the case of
> modules to a
> new framework the module might just as well get merged up to ease
> maintenance.

Once it's stable sure. If NTLM goes into head I have 0 objections to
merging it into auth_rewrite and leaving the separate ntlm branch for
experimentation (cf example domain membership based challenge
generation).

>
> New experimental developments/bugfixes might of course still be
> performed on the sub-branch even after the merging.

We're agreed.

>
> This probably also applies to the auth_rewrite / ntlm /digest
> branches,
> unless there are changes to the framework which you are not yet sure
> about.

The same logic applies. Yes I am happy to merge both the ntlm and
auth_digest branches up a level, leaving the old tags for experimental
work. My reference before about the ntlm tag was that IMO it is still
experimental.

> Having many separate branches containing tested code which do belong
> together the maintenance cost of the branches increases, and it gets a
> bit hard to test the things together as one unit.

For sure. I do find having to run a separate source tree to get
ntlm+digest a little annoying :]

Rob
Received on Fri Jan 05 2001 - 00:06:08 MST