Re: Best database for web backend on Linux?

In article <7076ob$suf$1@nnrp1.dejanews.com>, oliver.willandsen@sg.cec.be writes: >In article <i5AV1.200$Rj4.1148457@news.rdc1.sfba.home.com>, > "Yun Ye" <yye@home.com> wrote: >> Now that we got all sorts of commercial and free databases available for >> Linux, I don't know what to use. I currently use PostgreSQL, but its SQL >> implementation is quite incomplete (no foreign key, alter table add >> constraint, etc.), and there's no row-level locking. I want to try >> something else but don't have much experience with other databases. Can >> someone summarize the pros and cons of each database as a web backend? I'm >> considering PostgreSQL, MySQL, Solid, Oracle, Sybase, and Informix. The >> features I'm most interested in are transaction support, general SQL >> support, good Perl DBI/DBD implementation, row-level locking, online backup, >> max size of VARCHAR, and persistent connection. Oh, also price. (I know >> Sybase for Linux is free, but what's the price for Oracle as a web backend?) >> >> Thanks, >> Yun >> >> > >Both Sybase and Oracle have transaction support and row-level locking ( >Sybase only now ), Perl implementations exist for both - SybasePerl and >OraPerl, max size of varchar is bigger for Oracle, both ( as development >editions ) are free - Sybase will remain free, but without any support from >Sybase, Oracle as a deployment database will cost something similar to the >product on other platforms. > Sybase has a TEXT type, which is essentially unlimited in size. It also has the interested aspect of being stored on a separate page from the rest of the row's content. This is nice, in many cases, as it keeps the number of rows/page higher than it would otherwise be. If you're using VARCHAR/TEXT, though, you may be missing a factor. That is the "searchability" of these textual fields. I don't mean whether or not it is possible (of course it is), but how efficiently it can be done. In other words, is there a mechanism for doing fuzzy or partial matches (ie. like a glimpse db) that exploits indexing? There are engines (ie. excite or glimpse) that are good at full text searching. But these are tough to combine with the structure available in a relational database. Yet that is exactly what we need in many situations (and I doubt that we're alone). Also, if you're interested in atomicity and foreign key support, there's something else. Sybase only supports a "restrict" restriction on foreign keys (referenced rows cannot be removed). Missing are "set null" and "propogate delete". I don't know what Oracle supports in this matter. - Andrew