|
|||||||||||||||||||||||||||||||||||||||||||||||||||||||
News.Com: MSN outage enters fifth day. "MSN Messenger users started experiencing trouble Tuesday, with the most commonly reported glitches being connection problems and missing buddy lists of friends. Microsoft temporarily took MSN Messenger offline around 3 p.m. PDT Thursday in an attempt to fix the problem. Although the company had partially restored the service in many countries by Friday morning, at 10 a.m. PDT Saturday problems still plagued as many as one third of its users." Miguel de Icaza: "Recently, I have become addicted to reading your Scripting News and related sites." That's cool! Sean McGrath: How many XML gurus does it take to change a light bulb? Sometimes it works better if you don't bring the ideas to the XML gurus. That's been the experience with XML-RPC and SOAP, you can see the difference, one was frozen in 1998, and the other only stopped being a moving target in 2001. (Knock on wood.) They started in the same place but with XML-RPC we were able to get apps going immediately, although it took quite a few months to figure out how to do it. By the spring of 1999, while working on the first release of Manila, we did a web services preferences system, which unlike Microsoft's Passport, is distributed. We've been using it ever since. When you do a survey here you're using that preferences system, for example. I'm sure we're not the only ones with such a prefs system, for example all the Yahoo sites are tied together in a common membership system, but we did our work based on a public specification, XML-RPC, and now are ready to bring it into SOAP. Check out this spec, published on 8/11/99. "We're doing something new with preferences on UserLand.Com, we're distributing the database from a central machine where the user goes to change his or her prefs. We then XML-RPC the prefs to other servers ('affiliates') that generate pages so that user preferences can be applied to rendering the pages." Important note: We changed the implementation that's described in the spec, we no longer depend on a Batch Job Server. Instead we added asynchronous messaging to Frontier. Also, at the end of the doc I say we're going to do the same thing with story flow in My.UserLand. We tried it, and it scaled awfully, so we went to an all-static distributed approach in Radio UserLand, which works really well, no problems scaling, so far. (Praise Murphy.) We also implemented RPC handlers for membership in Frontier so that clients could be created in non-HTML environments. I really like the On This Day In feature more and more. Last year's Scripting News for 7/7 was really kickass, if I say so myself. 
|
 
|
||||||||||||||||||||||||||||||||||||||||||||||||||||||
|
© Copyright 1997-2005 Dave Winer. The picture at the top of the page may change from time to time. Previous graphics are archived. Previous/Next |
|||||||||||||||||||||||||||||||||||||||||||||||||||||||
