rest-discuss at Yahoo! Groups

Re: A different approach to supporting transactions over HTTP

Eric Newcomer — Thu, 07 Jan 2010 01:50:09 GMT

Compensation is not the same as undo, that is a big error in this discussion. Undo is a recovery mechanism. Compensation is something every transaction needs

Re: Rebooting REST

Eric J. Bowman — Wed, 06 Jan 2010 18:06:21 GMT

... I don't think that applying REST architecture extends REST in any way. "REST... [focuses] on the roles of components, the constraints upon their

Re: Rebooting REST

Eric J. Bowman — Wed, 06 Jan 2010 17:46:02 GMT

... I think it depends on what one means by fault-tolerant. I hate when a website hangs entirely because some ad rotator isn't responding. If my website

Re: A different approach to supporting transactions over HTTP

mike amundsen — Wed, 06 Jan 2010 14:42:37 GMT

Bob: I think once you are doing anything on HTTP it's all application-level. I've been focusing on the the control data aspect HTTP lately. It's

Re: A different approach to supporting transactions over HTTP

Bob Haugen — Wed, 06 Jan 2010 14:31:10 GMT

... I think once you are doing anything on HTTP it's all application-level. ... You may be confusing 2PC with ACID. Peter Furniss once told me that any

Re: A different approach to supporting transactions over HTTP

Bob Haugen — Wed, 06 Jan 2010 14:25:48 GMT

... P.S. compensation is also 2PC, just with an optional 2nd phase.

Re: A different approach to supporting transactions over HTTP

mike amundsen — Wed, 06 Jan 2010 14:14:51 GMT

Bob: Thanks for the feedback. I agree that my example does not cover all possibilities. In fact, I referenced (as "final commit") the interaction type you

Re: A different approach to supporting transactions over HTTP

Guilherme Silveira — Wed, 06 Jan 2010 13:31:16 GMT

... suggested, stuff happens... That's what I thought, but maybe I was wrong. Thanks Bob, Jan...

Re: A different approach to supporting transactions over HTTP

Bob Haugen — Wed, 06 Jan 2010 13:18:55 GMT

On Wed, Jan 6, 2010 at 7:00 AM, Guilherme Silveira ... If I understand correctly, you are posing a situation where the provisional phase (quotations) has

Re: Rebooting REST

Tim Williams — Wed, 06 Jan 2010 13:16:45 GMT

oops... +4s/isolation/reliability

Re: Rebooting REST

Tim Williams — Wed, 06 Jan 2010 13:07:13 GMT

... An "applied REST architecture" extends REST by definition, right? On the other hand, if you're talking about starting with REST and applying additional

Re: A different approach to supporting transactions over HTTP

Jan Algermissen — Wed, 06 Jan 2010 13:06:24 GMT

... This inevitably leads to 2PC kind of problems and the usually best way to deal with that in practice is to simply account for the fact that some errors

Re: A different approach to supporting transactions over HTTP

Guilherme Silveira — Wed, 06 Jan 2010 13:01:19 GMT

... (provisional) followed by an order (final). ... commit pattern, altho it does not obey all the ACID rules (which are impossible to follow RESTfully). I

Re: Rebooting REST

Jan Algermissen — Wed, 06 Jan 2010 13:00:07 GMT

... RESTs primary goal is not fault tolerance but decentralized evolution, scalability and simplicity. REST's focus is coordination of components owned by

Re: A different approach to supporting transactions over HTTP

Bob Haugen — Wed, 06 Jan 2010 12:45:22 GMT

... Compensation often does not work. Think of an offer to buy. If it's optimistically committed, both buyer and seller may take followup actions that could