October, 2003 Archive

Some hours ago my changeset number 1000 went into the xaraya repository.

As you can see in the comments, it’s not a good one, basically cleaning up some mess i made earlier that day.

Nevertheless, it’s there

Using a script from Bryan O’Sullivans bitkeeper repository we integrated the bitkeeper repositories for Xaraya with our bugzilla installation. The script is used as a trigger everytime changes are committed to the main repositories on the server and scans the changeset comments for text fragments mentioning bugs.

When it finds those fragments, the following happens:

• the change committed is attached as a patch to the bug; (this patch can in most cases be applied to source trees out there, while maintaining upgradeability)
• a comment is entered for the bug, summarizing what happened;
• links are created in the comment to view the path in the bitkeeper web interface;
• bugzilla mails the relevant parties as usual.

This script is smart enough, not to enter the same information on the same bug and changeset twice which, if you know how bitkeeper works, is not uncommon if you have multiple repositories. As always, on deploying something new, you always start to think immediately about features which you could add (example: scanning for fixed  or fix  and setting the state of the bug accordingly).

During the period of  little over a year, we’ve built ourselves an infrastructure which is still growing and is literally improved upon everyday. Some elements of it:

• automatic updates of mirror repositories;
• fully integrated patch system;
• custom command system, defining new commands in the repository which propagate to all developer repositories;
• funny stats output;
• getting information on RFC status;
• integrated build system based on phing;
• integrated source documentation generator based on phpdoc;
• repository web interface as module of xaraya;
• adding comments to files which propagate across repositories;
• integrated unit test system;
• semi-automatic archiving of old files in the repositories;
• possibility to test for compliance with coding standards on every commit automatically;

With hindsight, a rather long list of tools, mainly created for fun. Some are used extensively, others are only used by myself.

Worked a bit on the  blog by email stuff. Sanitized the script a bit. The posting should now be a bit more clean as the backend now only includes stuff between the body tags of the received html. Nowhere near rock-solid, but useable nevertheless.

Just for the technically inclined, this is the regexp used to extract the blog entry from the html:

!<body[^>]*>(.*?)</body>!

Next step, getting some regular expression which makes sense to extract excerpt and extended entry (if any) out of there. Thinking about using <hr /> as a separator; it just looks natural when composing the entry. Dunno.

As mentioned earlier i like to treat weblogs as much as possible as a messaging activity. Not because i like messaging so much, but it has become the major activity in working the internet for me; communicating through news and mail. I find blogging interesting, but the threshold was always a bit high.

Especially blogging on linux seems a tab bit less friendly than the excellent clients available for windows, and those didn’t fit my pattern. What I did was basically the following:

1. searched the net for email gateways to MT (found 3)
   
2. literally threw the three files into one perl script ( they all had something i needed)
3. reorganized and made it fit to be a baseline for me.

So, phase 1 is complete, i can post in html with my email client to my blog (through a special email-address linked to it), selecting multiple categories.(by specifying them in a X-Categories custom mail header). Attachments to the email are automatically placed in the right location.

Outstanding issues:

1. a bit better extracting of the message (i left this posting as it is generated, definitely not what you want for all postings, look at the source of this page!)
2. keywords, excerpt and extended entry support
3. being able to comment through the same mechanism
4. get the stylesheet of the blog as a template in the email composer (well, maybe not)
5. Editting of posts
6. specifying sites to ping

As I’m working on several different computers at different locations, with at least two different operating systems, it’s sometimes hard to maintain the status of all the message types you need to follow these days.

Mail, nntp, rss and websites all boil down to messages for me in some form, so I would like to treat them that way. It’s not easy to have all places you work behind a computer synced up. I got one step further tho with the help of nntp//rss

For mail only, my problem has been solved a while ago. Using the Cyrus mail server, i can access my mail from basically anywhere, maintaining status across locations and computers.

For newsgroups, the solution is basically the same, using shared folders in Cyrus as newsgroups and the ‘seen’status of the newsgroups is maintained as well. At this point, the complexity starts. I read newsgroups from different news-servers and chances are about 100% they are NOT Cyrus shared folders.

So, to be able to maintain status, I have to install a ‘proxy’news server which gathers the newsgroups from the different servers and then let the Cyrus server sync with that proxy news server. I’m not sure how posting to those newsgroups would work in this case, but i guess that is manageable by creating accounts in a clever way.

Now, for RSS-feeds i used to have a separate reader and couldn’t figure out why that was needed. I fed ‘rss to nntp’ into google and after some searching i found a couple of RSS to NNTP and RSS to IMAP gateways.

Installing nntp//rss was easy enough. It creates a newsgroup for every channel and behaves like a news-server for them, so you can read RSS feeds through your newsreader.

At this point i read mail, news and rss feeds all with the same program (in my case mozilla), the ‘seen’state is only maintained for mail, but at least the RSS subscriptions are all in one place, instead of on all the computers i work on.

For posting, nntp//rss allows you to define the defined newsgroups as ‘writeable’and link a blog to it using the blogger API or metaWebLog. Although pretty unstable and not quite useable yet, i really like the idea, especially if replying to a ‘newsgroup’message would show up as a comment on the linked weblog

In conclusion, integration is possible for me, because i run my own servers. The amount of work and the expertise needed to do it is, as of now, far too high for this to become mainstream.

Spent the most part of this day getting http links in thunderbird to open in firebird. Still not succeeding. This is pathetic, i’m pretty sure my local configuration is not broken. I’m NOT a linux newbie and can’t get it to work.

There’s no option to configure it, the system wide configuration in gnome is at least confusing, and no obvious way to debug it. On a related issue; had to install the MozEx extension and write a custom shell script to get mailto links to open the compose window in thunderbird when clicked on in firebird.

What are these people expecting from me?