Re: Mini code-sprint report

From: Henrik Nordstrom <hno@dont-contact.us>
Date: Thu, 27 Oct 2005 01:44:49 +0200 (CEST)

On Wed, 26 Oct 2005, Alex Rousskov wrote:

> If we are talking about applying a few existing stable conflict-free
> Squid3-ported patches to Squid2 and calling that Squid2.6, then it may
> be a good idea.

The suggested set is mostly conflict-free or where the conflicts have been
resolved already (or will get resolved within the intended timeframe
anyway).

> If we are talking about opening up Squid2 for adding "a very large list"
> of features that used to be coded (or perhaps just prototyped) at some
> point in time after the Squid2 freeze, then would not it be a better to
> make Squid3 stable instead? Especially if you think that all those
> features were ported to Squid3?

Most were developed and verified in Squid-2.5, and also ported to Squid-3
but most often without any significant verification in the porting.

> FWIW, when we try to get companies sponsor Squid3 work, the Squid2
> feature-freeze status is the single most important factor. If that
> status changes (whether officially or by public perception), it would be
> a lot more difficult to get Squid3 work sponsored (not to imply that it
> is easy now!).

For me the situation is somewhat different.

I have customers demanding certain Squid-2 extensions (or to be exact
extensions to the current STABLE release), and often more than one. As a
result I am most often working with something remebling the proposed 2.6
in one way or another.

Getting someone to sponsor development in Squid-3 is as you note extremely
difficult. But getting them to sponsor development to the current STABLE
release AND squid-3 tends to work out much better in my experience.

Getting someone to sponsor the efforts required to make Squid-3 STABLE is
not something I see likely. Even is hard get people motivated to
sponsoring the maintenance of the current STABLE release which comparably
is a much smaller job.

> Would it be possible to limit 2.6 changes to a select few, widely used,
> well-tested, Squid3-ported features unrelated to performance and similar
> optimizations?

It would. But I am not sure I would exclude performance optimizations.

Regards
Henrik
Received on Wed Oct 26 2005 - 17:44:56 MDT

This archive was generated by hypermail pre-2.1.9 : Tue Nov 01 2005 - 12:00:07 MST