Murthy Chintalapati wrote:
>
> [b]OPEN Support for concurrent versions on disk[/b]
> Given the compatibility guidelines from the upstream community, it's
> agreed that we should support having the provision to carry at most
> one one additional version for apache2 (i.e. provision for a 2.4.x
> version concurrently with 2.2.x). 

To be precise, let's separate the shipping/support policies from the
file system layout since these are distinct decisions (and one is
architectural while the other is more of a business/marketing decision).

There was agreement in the conference call to support a major.minor
versioned directory structure under /usr/apache2, meaning,
        /usr/apache2/2.2/...

Furthermore, after discussion about the binary compatibility of patch
versions (e.g. 2.2.4 to 2.2.6) there was agreement that iff it became
necessary due to an uncorrectable break in binary compatibility, a
future update might need to ship a major.minor.patch versioned
directory, but the intent is to avoid this.

(Whether any given OpenSolaris distribution concurrently ships or
makes available or supports one, two or ten versions under this layout
will be a "marketing" decision by that particular distribution.)

> It was also felt there would be no
> reason to have the complexity of multiple micro versions within a
> minor version (i.e. a future 2.2.8 release should replace
> 2.2.6). While the upstream community seem to provide the binary
> compatibility acorss such micro version releases, this project team
> will test the next micro release and participate in the upstream
> community to ensure such a binary compatibility is preserved.

To be precise, this community cannot ensure binary compatibility
because we do not intend to fork (this led to the backup plan of
major.minor.patch layout iff it becomes inevitable).  If any such
issues ever arise, Web Stack project will attempt to fix and push
upstream but the ultimate decision on binary compatibility rests on
the Apache community.

(I foresee maintaining a small set of local patches as necessary.
We'll need to evaluate such differences on a case by case basis. We
must remain compatible with the upstream version.)
        


-- 
Jyri J. Virkki - jyri.virkki at sun.com - Sun Microsystems

Reply via email to