Saturday, August 10, 2002

Oracle RAC on Linux: http://www.byte.com/documents/s=7519/byt1027639024201/0729_bar.html Moshe says that Oracle's RAC is cheaper than SMP, but reaches the conclusion that it's probably not worth the trouble. I'd like to know how much of a break Oracle gives you on CPU licenses - I wouldn't be surprised if it was more expensive. I'd argue that management overhead of sophisticated clustering is seldom worthwhile compared to just getting a bigger box with a warm spare. Moshe suggests kernel level clustering (like the kind he's developing) is the way to go. I'm not knowledgeable enough about it argue, but I'm dying to see how PostgresSQL and kernel level clustering get along. Real world examples of non-computational clusters are pretty scarce. 3:01:37 PM      comment []  trackback []

Dell No Longer Selling Systems w/o Microsoft OS [Slashdot] Would this news item be possible if state governments mandated an alternative operating system? Could Dell afford to opt out of those customers? 2:54:31 PM      comment []  trackback []

My Saturday rant I saw an article on news.com about a California ballot initiative mandating the purchasing of open source or free software, much like the bill put forth by the Peruvian Senator. I remember Ralph Nader proposed a few months ago to have the federal government seek a market rather than legal solution to the Microsoft monopoly by redirecting their IT spending to alternate platforms. The Register reports that the U.S. government is currently lobbying Peru to keep them dependent on Microsoft software, so putting forward this policy in the business-friendly Bush administration is pretty much a lost cause. However, I wonder why the 13 states who are still pursuing antitrust action against Microsoft don't pursue this strategy? You really only need one state to do it... then the case studies come out reporting how much money they saved, materials and training programs are developed to address retraining issues, etc. and pretty soon ever other state starts to think "Here we are in the middle of a recession. Does it make sense to slash our social programs to the bone while tripling the amount of money we are giving to one of the richest companies on the planet?" Norway, Germany, Peru, China, California, Mexico ... someone has to be the first to step away and categorically award monopoly software profits to free software services companies. Someone has to be the first to turn down the services and software bribes put forth by MS to remain dependent and give up their choice. I love the free market; Open source and free software play very happily in the free market. Let's help the market do it's job by correcting an imbalance. 1:57:34 PM      comment []  trackback []

OpenACS-on-MSSQL port milestone I've reached an important milestone in my MSSQL Server port of the openacs data model. The kernel and all unit tests compile and run. This means that I am reasonably confident that I can start using it in client projects. I'd like to figure out a way in which I don't have to hand-port the whole TCL API (which would not be fun.) Realistically, I don't need to do this, because for the limited data-model re-use scenarios I''m envisioning, rewriting the few procedures in Perl is practical. Not the site-node or apm type stuff, certainly, but the basic community framework is do-able on an as-needed basis. I wish the TCLScript project was further along: http://tclscript.sourceforge.net/. I could see that being a major shortcut, because I don't think porting the specific AOLServer-isms would be a big deal. I'm doing a lot of intranet-type of work where perlscript calls vbscript. I know it's not a paragon of efficiency, or a model for a high-traffic site, but it works and it's stable. I guess a more scalable (but somewhat inelegant) option would be to run an XML-RPC-serving instance of AOLServer behind IIS, which would support the community framework pieces that I needed. Exposing individual (hopefully stateless) API procedures wouldn't be too bad, and could be automated to a high degree. It would mean I'd have to tackle the xql files and the data access API, which is probably worth doing anyway. 11:14:32 AM      comment []  trackback []