| RE: [hyades-dev] The state of the HCE project |
I just got back from a week of vacation, and it appears a lot has happened during that week. I haven't read all of the messages on the mailing list, but I have read all in this thread. If it seems like I missed something, it may be that I have. That having been said, I personally do not think it is time to panic, but by all means lets talk about whatever concerns anyone has. I've felt for some time that I had a pretty good idea of the shape of what we were trying to create, and it seemed like what we were doing fit that shape. Perhaps not everyone shared the same vision. We should definitely talk about that. As a bit of background, I've attached the original proposal that I sent out in March that sent us down the current path. These were the Intel requirements, and in talking with Richard, it seemed like there was a fair amount of overlap with other groups he had spoken with. To this have been added various other concerns that have come up along the way and not been documented, but which we have attempted to account for in the design. My original intention was to modify the existing RAC as little as necessary. It seemed like the external protocol could be modified without having to change the core of the server code. However, one of the "concerns along the way" that I referred to above was Richard's feeling (if I have understood correctly) that the current RAC was a simple component that had been pushed beyond the limits of what it was originally meant to do. And so, since we were introducing a new protocol, this seemed like a good time to introduce some broader changes that would build in some growing room for future expansion. The message pipeline architecture grew out of our discussions in this area. It seems to me that it is in working in this message pipeline architecture that we have crossed the line from expanding the current RAC to essentially replacing it. I personally feel that this was a right decision as it puts us in a good position going forward. The specific requirements behind it were the need for a pluggable transport layer, and the possibility of supporting SOAP based protocols in the future (I'm not clear to what extent this second part is a hard requirement). As for what we are replacing, we are only replacing the low-level protocol directly between the Hyades client core and the server, and between the server and the agents. It has been the plan all along to maintain compatibility within the client interfaces so that existing client-side code would work with the new client talking to the new server. I'm not sure we've written that down anywhere. As for the agents, I don't think we've addressed that directly. There have been specific conversations about the overall transition plan, and I thought that it was agreed upon. Have I missed any specific areas that anyone wants to cover? -Andy -----Original Message----- From: hyades-dev-admin@xxxxxxxxxxx [mailto:hyades-dev-admin@xxxxxxxxxxx] On Behalf Of Allan K Pratt Sent: Wednesday, September 01, 2004 6:18 PM To: hyades-dev@xxxxxxxxxxx Subject: RE: [hyades-dev] The state of the HCE project I think Harm echoes my sentiments about not knowing what we're really about here when he says, "changes [should be] based on well identified requirements." I don't think we have established what those are. Vishnu wrote, "Though it is looking like house of cards..., at some point I feel that all these [documents] will merge and make sense." That's not how it's supposed to happen. The design and architecture shouldn't emerge from the implementation; it should be the other way around. The only time the ad-hoc approach works is when the people writing the command protocols already have a pretty good idea of the requirements and overall architecture in their heads, and it's the SAME idea, and they just haven't bothered to write it down. I think it's clear that the people contributing to these HCE document reviews do NOT all have the same design ideas in mind. Harm also reminds us that we can't be in the business of wholesale replacement of the Workbench - RAC - agent protocols and structures. As far as I can tell, that's EXACTLY the business we're in. Is there some kind of coexistence strategy? Are all the new commands and protocols being piggybacked on the current RAC as "custom commands" or something? If so, why? Is the RAC protocol really so awful that we can't accomplish what we want by extending it? Maybe it's time to step back and rethink this. We can start by answering this question: "What problems with the current Workbench/RAC/Agent system are we trying to solve?" Then we can ask, "If possible, how can we solve those problems in the context of the current system and not a new, parallel system?" -- Allan Pratt, apratt@xxxxxxxxxx Rational software division of IBM _______________________________________________ hyades-dev mailing list hyades-dev@xxxxxxxxxxx http://dev.eclipse.org/mailman/listinfo/hyades-dev
Attachment:
Hyades Protocol Proposal.doc
Description: Hyades Protocol Proposal.doc