Nat!'s Web Journal

• Added support for JDBC connection dictionaries. Probably I still need to deal with the custom value types (like 'c' and so forth).
• Fixed a bug with text/image types and Sybase 12.5.1
• assertConnectionDictionary tries to close the channel more desperately now
• attributes that have factory methods use improved logic for binding the sybase type to the input parameter
• Fixed the abuse of an NSAutoreleasePool feature (addition of nil object) via IMP
• Derived attributes are not disturbing the fetch anymore (should have been filtered by upper layer, but wasn't)
• fixed a problem in the capabilities parsing code, that manifests with unusual packet sizes. Sybase bug around.
• [NSDecimalNumber zero] problem cropped up again. Fixed hopefully forever by using cached copy of [NSDecimalNumber zero].
• fixed a problem in the cancel code, that manifested on 11.9.2
• isFetchInProgress adheres to the EOF documentation and does not report fetches in progress, if no rows are to be returned (but status/output parameters might be available)
• fixed -[MulleSQLExpression createTableStatementsForEntityGroup:] it now does not produce duplicate column names, also the order of the columns created by column name and not by attribute name anymore.
• Support for wide tables.
• Support for large char and varchar
• Support for unichar and univarchar. Note that because of some implementation details, unichar column results will always be "described" by the adaptor as univarchar.
• Collects multiple errors, so that more important error messages dont't get lost.
• Single character strings failed to retain properly (ugh!)
• Minor pretty up on warning message added in 1.6.
• Minor bugfix when a connection fails.
• Various optimizations
• If the Model is incorrect and an external type needs to be guessed, the adaptor is less likely to produce incorrect guesses, that lead to incomprehensible crashes later on. A warning message will be logged.

It's exactly what I want to connect my guitar with my Mac, but its not on the market. Now should I wait or not ? Has this product been cancelled ? Was it just a hoax ?
Argh! Firewire promises less latency than the competing USB adaptors and €150 would be an acceptable price.

DSL Modem

Then I moved.

I the new surroundings I had the same spatial problem. This time I decided to solve it wirelessly and to get rid of the noisy linux box. I wrote about this endeavour a little bit here in my weblog. Basically the setup was: DSL-Modem -> Router -> WirelessClient ..... Access Point -> Switch -> various machines. Though Trendware T-100 S4W1CA router doesn't have the most sophisticated firewall, it is all in all a good product. It worked. Almost. Unfortunately the D-Link was not too clever as a wireless client and needed to get a packet from the router first to know that the router was there to be happy and receive packets and to forward them to the router. It needed that initital packet from the router, because it didn't know the MAC address or the IP address of its client right from the start. In normal operation, the router never emitted a packet. Catch-22.
I still haven't figured out, how it eventually did work most of the time. I suspect some routing information package eventually brought things into sync. But it was a pain in the ass. So some of the time the wireless client was just ignoring packets and the internet connection was dead.

DSL Modem

The last month I decided to switch Internet provides from T-Online to 1&1. 1&1 had a nice package where I could purchase a DSL Modem + Router Switch + WLAN for - not so much money (I forgot how much it was). So I was going to combine the "DSL-Modem -> Router -> WirelessClient" threesome into one hopefully well working box.
I have to admit that this wasn't thought well through, because the "FritzBox" that appeared on my doorstep converted my setup rather to a "DSL-Modem ->Router->Access Point" scenario as it can't work as a wireless client.
Now that was truely a disappointment, because this obsoleted my DWL-700APs as they can't be used for bridging an ethernet based network to an access point. After a lot of head scratching I invested into a Linksys Wireless Ethernet Bridge WET54G. This device can be hooked up to an ethernet switch and is able to forward the traffic to the access point. Rumor has it, that it is limited to 32 devices. Documentation is really scarce. The setup screen is very barebones and there is very little in terms of status feedback.

Finally everything worked as expected. That was two days ago. Today for a final pretty up I decided that I should optimize one of my two power strips away, that I use to juice up the phone and DSL machinery. Obviously two outlets had become superflous. So I unplugged the FritzBox from strip A and plugged it into strip B. And now it can't do DSL anymore. It just broke. It can do wireless and everything, but the internal DSL modem has gone to bit heaven.

Currently I have again about 20 meters worth of ethernet cabling lying on the floor, that connect my LAN to the Internet over the Trendware router and the T-DSL modem.

So the lag (I think) is 1s / #samplerate * (samples_per_output). If the buffer is 128 samples long, the lag should be 1 / 44100 * 128 = 3 ms. Then the sounds travels to my ear currently at around 350 m/s. It's hot and humid here in Bochum. As my speakers are a bit more than 50 cm distant from me that should be around 2 ms lag. Interestingly one can observe this with Garageband or in this case Cacaphony a shareware Sampler. The left channel is the directly miced click from the metronome and the right (lower) channel is miced from a distance. Sound does indeed take some time to travel. Then my brain has to process the signal and tell my fingers to move. One can only imagine how much lag that means. Finally a note is sounded, the strings excite the pickups and a signal gets transmitted with a speed close to that of light (good ol' analog technology) to the line input of my G5. There it is sampled at a rate of 44.1Khz and sent to Garageband in small packets each. So this is the same situation in reverse, that I described at output time. So to measure the latency of my setup, I directly miced the speakers and recorded the result with Garageband again. As there was almost no air distance and no human brain involved, the latency would be incurred by the output and input buffers only. As you can see, there is no significant latency.

I wanted to record this with my iSight, but you can't use the iSight with Garageband. iSight samples with 48 Khz and Garageband needs 44.1 Khz. Pretty lame.

Homeless World Cup in Edinburgh - 17.07.2005 22:07
Fünf afrikanischen Teams die Einreise verweigert

Vom 20. bis 24. Juli findet in Edinburgh der Homeless World Cup statt. Es ist die dritte Auflage der Obdachlosen-Weltmeisterschaft nach der Premiere 2003 in Graz und den Spielen 2004 in Göteborg. Doch wenige Tage vor dem Startschuss bahnt sich nun ein Skandal an: Die Regierung Großbritanniens verweigerte fünf Teams aus Afrika die Einreise.

Die Teams aus Kenia, Burundi, Kamerun, Sambia und Nigeria dürfen laut einer Pressemitteilung des Veranstalters am Montag nicht nach Großbritannien einreisen. Als Grund wurde genannt, dass es Befürchtungen gibt, dass sich die fünf Teams während ihres Aufenthalts in Schottland nicht mit eigenen Mitteln selbst versorgen können. Die beiden anderen afrikanischen Mannschaften aus Namibia und Südafrika sind von den Maßnahmen nicht betroffen.

There is a new version of Grow available. It's as good as the old one and I solved it :)