- From: David Brownell <db@argon.Eng.Sun.COM>
- Date: Mon, 19 Oct 1998 19:44:13 -0400 (EDT)
- To: www-dom@w3.org
Catching up on some mail ... Keep in mind that the _real_ problem is accessing the data which is exposed by DOM, not accessing DOM itself. So it shouldn't be a requirement to have RPC calls going over the Internet that match what DOM uses; as folk have noted, locking and concurrency weren't exactly design considerations. I've encouraged folk to think in terms of shipping XML around, and reparsing as needed (perhaps into DOM). That means using what some have called "coarse grain" messaging. (Which is what many of us were trying to create with CORBA; but that's another story!) It's reasonable to have an HTTP request fetch part of a document, and let various web caches handle it, or let servers return the relevant fragment info by computing it dynamically. Then familiar APIs (like DOM) can be used to navigate. Another way to put this: networks do better at bulk data transfer (e.g. XML documents) than at low latency operations (like calling a procedure), and designs should accomodate that reality. Some of those document repository efforts (e.g. http://www.aiim.org/dma/) are worth looking at. - Dave
Received on Tuesday, 20 October 1998 13:20:38 UTC