Adding Transactions to Servlets with JOTM
Subject:   transactions and the web
Date:   2003-07-31 05:27:31
From:   anonymous2
I remember reading a long time ago about adding transactions to the web (may even have been before servlets). Took me a while to remember where, but how does this compare to:




Full Threads Newest First

Showing messages 1 through 6 of 6.

  • transactions and the web
    2003-07-31 06:19:28  jeff.mesnil [View]

    Conceptually, there is nothing really new, the differences are at a practical level.

    What has really changed since these papers ('97, '98) is how the web applications are developped:
    o the communication between the web and Java platform is now done by Servlets
    o the API to use Transaction is standard (JTA) instead of proprietary
    o Open Source software such as Tomcat and JOTM have commoditized and leveraged the use of these technologies

    If you're interested by new ways to think the relation between web and transactions, you may consider some of the web services transactions specifications which are out (e.g. WS-CAF which was designed by Arjuna among others)

    • transactions and the web
      2003-08-06 06:57:56  anonymous2 [View]

      What do you mean by proprietary interfaces? From my reading of those papers they are using a standards compliant transaction service. OK, it's based on the OTS not the JTA, but so what? It's not proprietary.

      BTW, I disagree think opensource has commoditized transactions. Tomcat is another matter entirely, but not transactions.

      Thanks for the other reference.

      • transactions and the web
        2003-08-07 01:50:31  jeff.mesnil [View]

        For proprietary interfaces, I was refering to the fisrt paper (Constructing Reliable Web Applications with Atomic Actions) where they define their own toolkit (W3OTrans). In their second paper , they're using OTS which is a standard API (I'd even say that is more standard than JTA). My mistake!

        As for opensource having commoditized transactions, I still think that it is true but only in the context of EJB. What we'd like to manage with JOTM is to bring the same ease of use of transactions in the context of Web applications.

        • transactions and the web
          2003-08-07 07:40:00  anonymous2 [View]

          I find your last paragraph interesting: isn't there already work going on in that space with JSR 156? Are you (or ObjectWeb) involved? Seems like you should be IMO.

          Also, what's your take on WS-CAF, WS-T and BTP then? Surely you're not advocating using the JOTM semantics you've described here for Web services transactions, when everyone else has basically said it's wrong?

          • transactions and the web
            2003-08-07 08:28:13  jeff.mesnil [View]

            When I was talking about Web application, I was thinking of JSP/Servlets specification.

            But, let's talk about Web Services applications.
            First we're not involved directly in JSR 156 (JCP and Open Source do not mix very well...) but we're interested to see where it is going.
            We believe that transactions processing with be a main issue when Web Services become more widely used.
            ACID transaction are not adapted to Web Services (they're short-lived and live in a trusted domain)
            So I don't think than ACID transactions will be a lot used for Web Services but instead WS will use Long Running Activities (or Business Transactions).
            As for our position wrt specifications (a similar discussion has been held on JOTM mailing list[1]): the choice is now between WS-CAF and WS-T (almost all companies behind BTP are now behind either WS-CAF or WS-T). We do have an implementation of BTP but only ACID transactions are supported.

            Companies behind WS-CAF have expressed their desire to give it to a open standards body (W3C or OASIS) and to allow Open Source implementation of it.
            WS-T is the property of IBM, Microsoft and BEA and
            it seems that an Open Source implementation of it is not legally possible.
            So our choice would be WS-CAF (*when* it is submitted to a standards body).

            hope it clarifies JOTM position,

            • transactions and the web
              2003-08-11 08:59:49  anonymous2 [View]

              Thanks Jeff, that clarifies things a bit. I think ACID transactions will be used where high-cost interactions are concerned. For example, if I'm transferring $millions I'd be happier if traditional ACID semantics were involved than transactions with compensation (if a failure occurs, I'd rather have my money back in my account now than have to wait for some lawyers to settle the dispute and get it returned from wherever it ended up - and compensations aren't guaranteed to complete anyway, so compensations may need compensations ad infinitum.)

              Also, the kind of ACID transaction models in WS-T and WS-CAF seem perfect for interoperability between traditional TP systems. Something even the OMG failed to be able to pull off!

              I think I understand where you are with regard to WS-CAF versus WS-T and I agree.