From: Jim.Johnson@software-exploration.nospam.com
Sent: Saturday, November 04, 2000 12:17 PM
To: Info-VAX@Mvb.Saic.Com
Subject: DECdtm, XA, TIP meeting report

I've just gotten back from my week in the US, during which I met with
VMS Development to discuss what the situation was wrt DECdtm, XA, and
TIP support.

For those of you who may not have been following the DECdtm topic, a
little background:

A couple of months ago I learned through this newsgroup that VMS had
apparently decided that it was too difficult to add XA or TIP support
to DECdtm.  As the original designer of those services, this viewpoint
surprised me (to say the least).  I was asked in this newsgroup what
it would take for me, or my company, to add such support in.  I
remember replying that Compaq would need to pay us ... and from that
statement everything else has sort of wound its way to here.

Let me say, as a quick aside, that I hold very strong opinions on the
value of transactions as one of the more powerful models for reliable
distributed processing, and that I've never been reticent about
voicing that opinion -- either before or after leaving the VMS
development group.

Anyway, I opened up contact with VMS development over this topic, and
arranged to see the people there who are responsible for setting the
direction and agenda for TP in VMS.  My goal was to find out just what
was happening, what decisions had been made, and whether or not I, or
my company, could help them in reaching a technically and commercially
solid state.  The remainder of this note is my report to this
readership on what I found.  


I met with the individual who is responsible for establishing/updating
the VMS TP strategy.  This is someone who I have known and trusted for
years.  Our meeting overran considerably, and covered what they were
doing, the state of their plans, and how they arrived where they were.
It was a good, open, and informative meeting.

The essential news is that VMS has what I would consider the correct
basic goal -- namely to support interoperation with other OS and
application platforms with minimal changes to any existing software.
What I was shown indicated a level of interest in this topic that is
the highest I've seen it in a decade.  Furthermore, the declarations
that I've read here that RTR will be the basis for this vision are
premature.  While there may have been a view at one time that RTR
would be the basis for all future work, that was not the current view
as I had it expressed to me.

Now, RTR clearly does have a role in any overall VMS TP strategy, and
I personally fully expect to see further work done on RTR.  However,
to reiterate, I saw nothing that said that VMS had decided that all
future enhancements would be via RTR.

Let me be clear -- if you believe that VMS should support XA and TIP
without requiring an RTR infrastructure (say, perhaps, via DECdtm),
this was *all* good news.

However, the flip side of this is that the current state is partly
because VMS has not solidified its plans for how to achieve the level
of transparent interoperability that it feels it should provide.
Although I've been known to hold cynical views on plans such as these,
I have to say that what I saw and heard suggested that VMS was finally
treating this difficult issue with the level of seriousness that it
has long deserved.  The fact that they haven't knee-jerked a response
is a good sign.

This is the core of the two possible negatives: will the features show
up in time for you, the users, and will the plans end up requiring RTR
after all.  I can't answer either of these.  I'm worried about the
former (albeit with no data), and hopeful on the latter.

Finally, I was asked to consider a reasonably significant
participatory role in establishing this TP direction.  This is
something we've agreed in principle, and are currently pursuing.



What I saw indicated that it was understood that this was an important
area for VMS.  However, I believe that there are two things that
people who believe this to be important should still do:

1. let VMS know that it's important.  While I don't worry just now
that they don't believe it to be important, continued input will help
keep minds focused.

2. let VMS know just what you want to do.  This is actually the much
more important issue.  What are you trying/wanting to do with TP and
VMS?  Who are you trying to interoperate with, and in what way?  Is
COM important?  Is Java (J2EE)?  What about other TP monitors, other
databases?  There are lots of possibilities, and the technical issues
may stretch well beyond the 2PC protocols.  The more known about, say,
which vendors and systems are vital, the easier it is to focus the
work and plan for an effective rollout.

These things will help, and I'd ask anyone who is interested in having
something happen to TP on VMS let VMS know.  I'd also ask that you let
me know.


As a final postscript, I spoke to the people who could be considered
"the VMS Brains trust" and none of them claimed to have made the
statement about DECdtm's upgradability.  Their thoughts were that
either a) it was taken out of context, or b) someone was speaking with
an authority they didn't have.  I think we'd all like to know which,
so if anyone can get me details of where they heard that message from
I'd certainly appreciate it.

Jim.

Jim Johnson
Software Exploration, Ltd.
Software Navigation and Discovery Tools