Older blog entries for pjones (starting at number 123)

Perl's DBI API sucks
Python's DBI API sucks

Oracle has the right idea for transactions; a transaction is a child of a connection, and a statement is a child of that. Perl and Python both assume that a connection is a transaction, and that really sucks.

Oracle sucks

OCIServerAttach() allocates some memory; its reciprocal function, OCIServerDetach(), frees 344 bytes less. That's right -- if you make a connection to Oracle, you're excersizing a memory leak.

So I wonder what it takes to get mirwin to certify you.

katzj: lots of questions still need answering though; not the least of which being intermediate bytecode execution and write-trapping. Oh, and deletion.

katzj: probably not. I think I told you the gest of it at chargrill once though.

It occurs to me that I'm having ideas so bad that they just might work.

mbp: but what about yogurt?

/me agrees with cturner.

In other news, I finished reading my second computer book :). Damn, ADD is a pain in the ass. Anyway, the book was See MIPS Run which I recommend photocopying; the author isn't very nice. Next in line: the dragon book.

pictures

New ones are here and here.

work

There is no ROSWELL beta

Packaging is hard. Lets go shopping!

Discovered that there's really no good way to match a package up with its source package if you've got that package on more than one arch. Suck.

I'm missing OLS because I'm heavily involved in the next RHN release, and we're in the painful grind point of the project. *sigh*.

In other news, I've got a prototype for generating sql (possibly of any variety) from a live oracle database, and diffs between two db's probably isn't very far off. So I might be able to do "schemaDiff oracle_db pg_db" and get some postgresql sql which makes the schema look alike. Obviously, data convesion is still a manual process, as are store procedures (including check constraints :/), but its a start.

It also shouldn't be too hard to generate the SQL schema from any other db type.