home *** CD-ROM | disk | FTP | other *** search
- Newsgroups: comp.databases
- Path: sparky!uunet!spool.mu.edu!darwin.sura.net!zaphod.mps.ohio-state.edu!sdd.hp.com!scd.hp.com!hpscdm!cupnews0.cup.hp.com!dhepner
- From: dhepner@cup.hp.com (Dan Hepner)
- Subject: Re: Need help with X/Open Distributed Transaction Processing Specs
- Sender: news@cupnews0.cup.hp.com
- Message-ID: <BxGzA9.AFt@cup.hp.com>
- Date: Mon, 9 Nov 1992 22:18:09 GMT
- References: <1992Nov9.212407.17572@cs.cornell.edu>
- Organization: Hewlett-Packard
- X-Newsreader: Tin 1.1scd1 PL4
- Lines: 35
-
- From: jain@cs.cornell.edu (Neel K. Jain)
-
- >The XA specifications give me the impression that the application
- >program, the transaction manager and all resource managers (or their
- >stubs) are statically linked. Is this really the case or am I missing
- >the point (won't be the first time)?
-
- It really is the case that the model describes interfaces within
- a single process, but easy on that "statically".
-
- This single process view does not however preclude any of the parts,
- including the AP, from utilizing other processes in accomplishing
- their goals. It's just that there is a single process which contains
- the linkage of all; this process is the subject of the X/Open DTP model.
-
- One might wonder what is distributed about that. Nothing is.
- Using standardized transactional communication, two APs, and
- thus two instances of the model can participate in a distributed
- transaction. These are as yet unpublished, but are clearly
- the next step.
-
- There is nothing precluding dynamic linking however. Some "dynamically
- registering RMs" may well be linked, on demand, at run time; the essence
- is that they fit the notion of the model as focused on a single process
- where the interfaces are intra-process call based.
-
- >Also, has anyone used their reference model publication? Is it any
- >better than the specs?
- >Neel Jain
-
- The Reference Model attempts to answer your question, while the XA
- spec's primary purpose is to offer the details of the TM <-> RM interface,
- while referencing the Reference Model for the overall picture.
-
- Dan Hepner
-