Seb's weblog

So, I was just fooling around with google’s geocoding API and doing tests with different target addresses. And suddenly, my test code barfed about malformed xml.

A quick look at the xml file that was printed on the console didn’t indicate anything obvious. One thing that could indicate the source of the problem was that the “missing” xml tag was preceded by an accented character.

Since when is my console UTF-8 aware? And the XML declaration says this is UTF-8. What would happen, if I put an accented character (not UTF-8 encoded) before a < ? The parser would try to expand the character with the following one, hence eating the < and breaking the tag, and this breaks client code.

Sure, as a quick search revealed, I can add oe=UTF-8 at the end of my query string (or I could convert the output on my side before parsing, but no, thanks!). BUT WHY THE HELL ARE YOU DECLARING UTF-8 IF YOU’RE NOT REALLY SENDING UTF-8?!@# Makes me wonder how their xml generator is written. A VB script using some kind of printf? Toh!

So currently, my dumbest geocode API bugs:

• the server lies when it says it is sending UTF-8

• the server can’t handle POSTs (Is it that hard to do both GET and POST? Don’t answer, please…)

Another victory of the French IT representation: Pages Jaunes launched its new web site, more modern and all. Apparently, what they’re trying to say is that they can find other things than just phone numbers, and it should be easier now.

But have they looked at recent (or less recent even) pages done for searching and just searching? Google’s (basic) homepage is less than 20kB. live.com’s homepage is 30kB. The new pagesjaunes.fr homepage is weighting 200kB and I’m not even counting the flash animations. There’s almost 110kB of javascript? Oh, sure, I can see that this is well documented (and in French, nonetheless), but is this really necessary on a production homepage?

What can we find on this homepage? Two search fields (well, ok). Three Ads (two in flash) and some random informations (well, not really informations, just a flash animation for searching things about cities, oh well…). All of this for more than 200kB. Some would argue that for example Yahoo!‘s homepage weight more than 300kB. But two things have be taken into account for this: first, there’s content. Two, Yahoo! can afford it. Why? Because they have large pipes, redundancy and compute power. It took me probably less than a second to load the full homepage on a modern computer and with a decent connection, even though my cache was empty. On the other side, I tried 2 or 3 times to load pagesjaunes.fr when it took more than 20 seconds to display anything.

And don’t you tell me about the announcement-effect or I’ll hurt a baby rabbit.

I think Google will become my first choice for finding business numbers, even though the local informations for France are still pretty rare.

editing xml configuration items encoded in base64 and stored in a ldap tree. Everything is A-OK.

Maybe I should update my version of Glassfish, since I’m still using the V1. But however, there are a few interesting “bugs”.

Today, I won’t talk about my fabulous “lock-me-out-after-deploying-anything”.

But in this version, there are a few self-destructive actions that are allowed and authorized, although, well, they make the server unusable, not able to start again thereafter. For example, changing the IP for the JMX service: stop the server, start it… whoops, won’t start. Deleting the JMS service: idem (but I can change the reference IP).
I want to reduce the number of services of my instances and I want to have the same TCP ports for every one, on different IP addresses on the same system. and JMX is the only service that won’t allow me this.

Impossible d'accéder à un affichage utilisable sur le système distant. 
Voulez-vous continuer en mode ligne de commande ? (oui/non)
non
Not a valid answer. Please enter y/n:

I love such serious software that has been tested as any enterprise serious software should be.

I think that now I know why I prefer Eclipse. Just to start, if I create a new Java class, with Eclipse, I can specify:

• which class it inherits

• which interfaces it implements

Why aren’t those simple tasks available with Netbeans?

Only thing that I “prefer” in Netbeans compared to Eclipse, is that I can checkout a project before actually creating the project (especially considering that project informations or ant build files or the directory tree might not be available otherwise). But I create more than one class per project, so…

Short list:

• Netbeans support of SVN (bleh!);
• Eclipse update system asking again and again and again and again the preferred update site;

• WordPress support for multi-homed host. And breaking the freaking URL withouth checking, at that!

I can read my news feeds, my comics, my friends’ weblogs through Google Reader. But I can’t search in my subscribed feeds?! What the…

...or how happiness can lead to frustration.

So, a few days ago, I had watched a video found on google and tonight, I wanted to see it again. So I fire up a few searches that I thought would lead me to it, but no matches. I can’t find it. I can’t look up my browser history, since I’m not on the same computer (and the other one is sleeping or not finding its network, whatever!). But being a merry camper as I am, not afraid of salt mines or something, so being remembered that there’s a search history into google, I look it up and after a few seconds I find my link back.

Everything is fine, then? Almost. Except that I can’t do anything without having to use local bookmarks. I use google’s notebook which offers me a note this link for my web searches — but it doesn’t appear for video searches. And anyway, it would require a browser extension — which doesn’t exist for Safari.

On the search history page, I have a list of google’s search engines, my google bookmarks labels, the list of results and a calendar. “Oh, if there are my bookmarks, maybe I would be able to add directly a result to those?” No! The only proposed way is to add a scriptlet that would open a new window (once the target opened) to add the bookmark. Why can’t I just say “add this search result to my bookmarks” ? Doesn’t this sound wrong: “Bookmark pages more easily. Drag this bookmarklet…” ? I always thought that having a “click here” link on a web page just meant something was wrong with the UI, and the bookmarklet references just does sound the same to me.

And is there somehow a “Add features, but don’t integrate them” rule somewhere?

Gargl…

writing a presentation, I used to copy/paste code from xcode to get some nice formatting. I need some weird language, so I try to do it with Textmate. But copy/paste won’t keep the color formatting. test: FAILED. (emacs won’t do it either, but that doesn’t surprise me much).