[os-infrastructure] Re: [os-engineering] Model discussion boileddown

[Joe Abbate]

Hi Steve,

Stephen Ball wrote:
> Remember to keep a balanced perspective here. What we are deciding is
> not "do we want to support VMS", but, "do we want to allow the community
> to conveniently build the code and develop on VMS by directly using our
> community facing code management system".

But that was exactly my point in my earlier reply to Andrew.  Is there a 
VMS community that is interested in building Ingres from source?  In my 
experience, only one customer and one non-customer raised their hands 
and that was three years ago.  More importantly, if there are some 
customers, paying or not, that are interested in Ingres on VMS, do they 
see it as a long-term platform considering HP's announced plans for it, 
and if they do, what features are they most interested in?  Again, in my 
experience, VMS customers have expressed interest in (a) clusters, (b) 
OS/POSIX threads (which has now translated into Itanium support) and (c) 
large memory (P2 space) support.

It is possible that some customer staff may be interested in improving 
some aspect of Ingres other than these three.  If that's so, it may be 
either a generic or a VMS-specific feature.  In the first case, they 
most likely could do the work much more easily on Windows, Linux or 
OS/X.  It's only those in the second group that could want a VMS-based 
build environment--and one could go even further and constrain that 
group further since, for example, to modify and test a DCL script all 
you need is an editor and an Ingres installation.

> .the second statement seems a
> little less drastic to me, and I'm assuming, although less convenient,
> there are ways one could use VMS to build and develop, and use a proxy
> of some kind to interface with Subversion. If we decide we do want to
> allow the community the convenience of direct access to our code
> management system from VMS, the we still have the option of using
> Subversion, but we would have to do the VMS port ourselves, all this
> would have to be weighed against the implications of not using
> Subversion (presumably the only good VMS option is Mecurial).
>   

Although Mercurial may be a possibility on VMS, and from I've read it 
supports import/export from/to Subversion, because it's Python-based, it 
does have a number of modules written in C, so it remains to be seen 
whether it's an option.

> My current inclination would be to use Subversion and document how VMS
> developers could work through some code management proxy.
>
> Other potential options for VMS might be to have some kind of
> WebDAV/DeltaV client (does VMS support that), or use some Java port of
> the Subversion client. 
>
> None of the options are optimal or very convenient, but in my opinion
> they are better than not using Subversion.
>   

I still maintain that deciding whether to support any platform (whether 
it's VMS, Unixware, Tru64, FreeBSD, etc.) in the Community Edition, 
should be primarily a business decision (just as we decide what to 
support on Enterprise).  On the one hand, Ingres Corp has to consider 
its paying customer base and its staff and other resources.  On the 
other hand, the community itself should provide input.  We know they 
want Linux, OS/X and (I presume) Windows.  Maybe Roy Hahn, Karl and 
others should canvas the community, via comp.databases.ingres, the 
forums or email, to get that feedback.

Joe