Older blog entries for camber (starting at number 4)

Been doing alot of work on MDB Tools lately. I actually have a SQL engine up and running! This is my first big lex/yacc project, and I'm quite pleased with how easy they make life once you learn them. A few problems here and there but nothing big.

Anyway, finally figured out how Access stores its indexes. Not pretty at all!

On other fronts the dying corpse of Northpoint has left me cursed to dialup. My ISP tried to re-apportion the line to another DSL carrier but no one else comes out this far...argh. So, next saturday I get a cable modem. I hate the cable company with a vengence. They're charging me $10 more a month because I dropped their sorry cable service two years ago and got a dish.

Finally strung CAT5 to the basement and bought one of those linksys DSL routers (Yeah I know I could do it with a spare linux box, but I didn't really want a box on upstairs 24x7, and her office where the DSL line comes in is crowded already). The linksys is working pretty nicely.

I'm getting about 620kbps transfer rate off the supposedly 784k DSL line, but by the time you factor in IP overhead it's probably about right.

Released FreeTDS 0.51 this week. So far so good. Need to automagically find the 16 and 32 bit types for the platform it's built on. inttypes.h provides them on many platforms, but not all and where they are defined seems to vary a bit. glib using some sizeof macros to do it, so I'm probably going to follow that unless anyone has a better idea.

Also been working on a TDS based connection pool program. It works like this. Let's say you have a 10 CAL license for MS SQL Server and want to use Apache/PHP to pull data from the server. Now most likely you'll have many apache children (in a heavy traffic site anyway), more than the number of CALs you have. And more often than not they will not be querying the database at any particular time. So what to do? buck up for a 50 CAL license? pricey.

The TDS Connection Pool sits between the Database and the PHP (or whaterver) clients. It acts like a MS SQL Server, a client logs in and gets a forged login acknowledgement and thinks it connected. Meanwhile the pool maintains a group of logins to the real dataserver and when a client sends a query, it allocates a connection, forwards the query, forwards the response back to the client and then returns the connection to the pool. This has the additional advantage that the client-to-pool connection is very quick compared to the long connect time to the real dataserver, so you get the advantage of persistant connections without the disadvantages of having one connection per Apache child like PHP.

Working on FreeTDS bugs...Microsoft decided under SQL Server 7 to change the default packet size which has caused all kinds of havoc with large querys. I'm glad to finally have a handle on why though.

Been comtemplating how best to debug with a complex system like apache+PHP+FreeTDS+SQL Server, Problem is when you have 30 httpd children running around and you have an intermittant bug it's damn near impossible to catch it. Best I've been able to do is generate tons of logs and hope something gives.

Skipping the presidential debates tonight, think I've seen enough in the first two. (Go Nader!)

Generally hard to work on much for the past few weeks. I am buying a house, and if that's not bad enough the particular house we are trying to buy is in bankruptcy. I've never dealt with so much paper work in my life!!! argh! Anyway, this has really taken its toll on my hacking time.

I've been trying to get the stubs in place for QtK so I can get SybSQL to compile. It's pretty slow going though.

On other fronts, Comdex was pretty cool. It really struck me when I walked in and saw Linux everywhere. No blank stares with I asked for linux support from SAN vendors and such (looking at SANs for my day job). Any lingering doubt in my mind about Linux being mainstream have been dispelled.

Veritas had the coolest hand outs, all their people were dressed in doctors scrubs (something to do with data recovery I guess :-) Anyway, they had prescription bottles filled with jelly beans.

Redhat and VA were noticably absent. Corel had a big booth, and if I didn't know better I'd have sworn they were demo'ing Windows software! Not sure how good a thing that is.

I had a theory about the correlation of a company's financial situation and how scantily dressed their sales women were. Pick systems has apparently seen better days.

Looking through the freshmeat entries to put them on here, I just realized that I've started a new project every January for the last three years. Ah, January when the sun goes down at 5pm and the only thing to do is hack. No wonder the scandenavians are all such awesome hackers!

Anyway, been working on QtK again after a long hiatus. QtK is a really foolish attempt of mine to run programs written for Qt on gtk+ instead. It's coming along pretty well though. I've chosen SybSQL (a graphical Sybase interface written with Qt and dblib) as a porting target and have it almost compiling. Next comes the hard part!

Merged in some more contributed patches for mdbtools. Carl got some foreign key stuff working on the schema export side of things.

I'm doing a little talk at my LUG meeting this month on FreeTDS and MDB Tools, and I still don't have any material ready!