Now, after about an year of operations, it’s becoming evident that most of us find it cumbersome, at best. This wouldn’t be a problem by itself, if only it didn’t cause a slight resistance at the hour of using it; which is a problem instead.

Don’t get me wrong, I think SourceForge is a great place to be hosted on, and they have built an impressive infrastructure for OS projects. I love it.

The problem is that the interface is struggling. Slow to use, it feels like being in the 90s: it takes way to long for someone which is only moderately interested in reporting a bug to actually do it. From a maintanier point of view instead, it’s hard to get a bird’s eye view (does this rings you any bell?) over the overall status of the project (*).

Having seen or used a number of alternatives we have come up with a list of alternatives among which pick our new tracker. Being an OS project our requirements is just for it to be either OS itself, to be free to use by OS projects, or to provide an OS license. On the feature side a plus would be to be able to import our current ticket database from SF.

Currently our list is made of

• Atlassian JIRA
• Trac
• Launchpad

Now, here goes a digression about each of them

JIRA

If you have seen it, and most of you are likely to have done so, you’ll agree that it’s massive. Powerful yet reasonably easy to operate. It’s a commercial product which provide an OS license which would do the job. Furthermore Atlassian is well respected among the OS community.

Also, it seems that it’s possible to import SF’s tracker backup into it, although it’s not an official feature. Nevermind, as long as it works.

Trac

An OS project, which I must say I haven’t used seriously. Nevertheless it has an important feature that doesn’t ship with JIRA, a code browser. I wonder why Atalassian doesn’t do something like that…ah no, they do indeed:)
Beside that it also provides a wiki, which we have already. We might consider migrating it as well but not at first.

On the downside the setup is more complex than for JIRA and it has some more restrictive requirements. Our server setup supports them but we might move to an hosted service which might not.

Launchpad

Ok, this is not really a fair comparison, as we should compare the whole SourceForge hosting with it, but for the sake of this post I’ll limit myself to their tracker.

First of all, it’s not an issue tracker, it’s a bug tracker. Launchpad supports feature tracking through a Blueprints section, which supports task tracking and workflows though. Ok, we don’t use workflows for our tickets here:)

The UI feels much like JIRA, I haven’t tested it if not for browsing others’ tickets. It has the big advantage for us of being an hosted instance.

Launchpad itself might need a post itself. I like their approach but I’d need to document a bit more over their vision, as if we’re moving I’d like for our new place to be still there in a few years time at least.

Roundup

You might have noticed that there is a notable absent in this list. I haven’t included Google’s code for various reasons. It’s strange to say this while being a GMail addict but I just don’t feel like hosting Atalaya on Google. I know they are already harvesting the same information but still. Call me paranoid.

I don’t think we’ll make a decision during this post. Here I’ve just aimed at writing down some pros and cons that I’ve found for each option.

(*) These arguments are in fact not limited to the tracker, similar arguments can be used for release management, but for the sake of this post I’ll just talk about this piece.

My macbook is running the vanilla apache2 installation that Apple provides, and shipping with php 4.4.8. Apart from this I had already installed MySQL 5.0.27.

Running the installer has given me a bit of an headache though, as I was constantly getting an “Unable to connecto to database” error. Googling a bit had reminded me of all those problems people are having with the new authentication introduced since MySQL 4.1 (remember the OLD_PASSWORD thing?).

Nevertheless my problem was another: it happens that the default PHP module looks for the mysql socket in the wrong path. Setting


mysql.default_socket = /tmp/mysql.sock


in the php.ini solved it (note that the socket location may vary, check your MySQL installation).

Maybe this helps someone.

One of the things that kept us busy in the past month has been this bug. Basically we were rendering awfully on IE.

We’d had this issue for a while and it was just time to get dirty on it. Working with Jan has been really instructive and, beside reminding me that developers suck at designing UIs, I’ve learnt that supporting IE6 it’s not really a matter of throwing fixes or using javascript libraries reporting to support it (we use JQuery here).

We could get the basic layout fixed, but we didn’t get anywhere near solving all the issues we had so, after two days spent on black magic css-ing, we took the decision: Atalaya (at least in its 1.x versions) won’t support IE6.

IE6 has still 30% of browser share, but it’s been constantly declining over the past here and especially after IE7 was released. Also, our target users are probably the least likely to have IE6 on their corporate computers, which makes this decision reasonable after all.

This decision aligns with our objective of taking a fresh look at our UI for Atalaya 2.0 (hold on:)), thus not spending a lot of time on the current look&feel (yes, the JFreeChart charts really suck we know).

Ah, the online demo has been updated with the fixes, and contains some new features also. Go check it out:)

I’ve finally decided to post another no-brainer JQuery plugin: a simple news box fed by an RSS feed. You can see an example on our project’s home page. I had found the idea on the JQuery mailing list but loving plugins (although not really good at it) I decided to wrap it up a bit.

The result is quite simple: just place a div somewhere and initialize it with a feed URL:

<script language="JavaScript" type="text/javascript">
$('#rightcolumn').feedreader({
targeturl: 'http://blogs.atalayasec.org/atalaya/?feed=rss2'
});
</script>
...
<div id="rightcolumn">
</div>

There are two other configuration options available; how many posts to put in the news box and the length of the text to be displayed for each item. A complete initialization would be then like the following:

$('#rightcolumn').feedreader({
targeturl: 'http://blogs.atalayasec.org/atalaya/?feed=rss2',
items: 3,
descLength: 15
});

The code is pretty rough (and not really js-style) but functional. In case it’s useful to someone.

The original code, from Stoyan Stefanov, is here. I’ve simply wrapped it in a jquery plugin. It’s enough to wrap the canvas and table elements in a div:

<div id="pie">
	  <table id="mydata">
	    <tr><th>Lang</th><th>Value</th</tr>
	    <tr><td>JavaScript</td><td>100</td></tr>
	    <tr><td>CSS</td><td>200</td></tr>
	    <tr><td>HTML</td><td>300</td></tr>
	    <tr><td>PHP</td><td>50</td></tr>
	    <tr><td>MySQL</td><td>30</td></tr>
	    <tr><td>Apache</td><td>10</td></tr>
	    <tr><td>Linux</td><td>30</td></tr>
	  </table>
	  <canvas id="canvas" width="300" height="300"></canvas>
</div>

and to initialize the plugin:

 $(function() {
         $(‘#pie‘).canvaspie({table: ‘#mydata‘,canvas: ‘#canvas‘})
 });

The plugin hides the table and shows the pie. An example (and the plugin itself) can be found here, in case someone finds it useful.

P.S: yes, I do need a code highlighter, sorry.

P.P.S: Thanks Stoyan for pointing me to hiliteme.

Update: I’ve updated the test page to use ExplorerCanvas, so now the test page should work on IE as well.

• it seems that the CIS benchmarks are no longer supported, basically provided as is
• the output format binds to the xccdf-1.0.xsd schema, but actually contains constants defined in theresults, that belong to the xccdf-1.1.xsdversion.

About the Castor problem I had yesterday: the XCCDF schema has in fact a duplicate attribute in its segment definition, causing the default Castor configuration to bail out.After posting on the castor mailing list I’ve found the solution (thanks Werner): write a custom binding file.

In the meantime I had worked out an Xmlbeans version of the parser, so I think I’ll stick with it to implement a first working version of the CisNg appliance.

After reading this post it also seems that implementing such a map should be feasible. Instead of XmlBeans we are trying to use Castor for generating the object model, as we really need just data binding, not fiddling with the document itself.

Nevertheless it seems that the XCCDF schema (here) has a sneaky bit

Now, these two attributes will cause Castor to complain like this when generating the object model:

duplicate name found: _id

I have yet to see if there is a way around this, as I’m not really into Castor. I suppose that I need to specify a custom mapping through XPath…

p.s: Castor needs also imported definitions to be present locally. In this case I needed to fetch these files:

• XML definitions: XmlSchema.dtd , xml.xsd, datatypes.dtd
• simpledc20021212.xsd
• platform-0.2.3.xsd