Call Me Al

Looks like Blogger has a handy rich-txt editor all hooked up. A few small bugs, but so far it seems to be working in Firefox branch nightlies. I take that back. In Compose Mode it keeps triggering Find As You Type. But that could be the hokiness of the new FAYT bar, or some sort of focus issue. I won't blame it entirely on Blogger quite yet. However, it now checks for unclosed tags!

I've been having trouble understanding how install.rdf, update.rdf, and the internal extension compatibility setting all go together. Ben Goodger was kind enough to spend some time on IRC with me and step me through the process. Although he is working on documenting the process, I'll sum up what I've learned. The internal compatibility setting is basically major.minor. The only time that will change is if something was changed for a subminor release and it breaks extensions. That should be very rare. This internal version will tell Firefox that an extension for 0.9 should work on 0.9.2, for example.

If you haven't noticed, updates to your themes and extensions on UMO aren't being detected by your browser. It sounds like the web service will be deployed soon. Then again, I've heard that before.

In order to make update.mozilla.org useful to developers, it will need to be easy for them to add and update their extensions. Since the site has gone live, I've added/updated more than 50 myself. What if you create an extension that works in Firefox 0.9 and then 1.0 comes out? Right now that means that an extension developer needs to update install.rdf, submit it to Bugzilla to be uploaded to UMO, and wait. Eventually the site will be self-service for authors so that they can upload themselves, but that still means updating the extension for each release and uploading it. That maxVersion is a limiting factor. Well, we could make maxVersion optional. But then if something changes in Firefox, your extension could mess up the build. The same goes for setting an artificially high maxVersion. While that may be fine for a corporate deployment, where things are tested, it has the same problem. How do we get around maxVersion in install.rdf? What if Firefox checked update.rdf to see if maxVersion changed? When an extension is being installed, first it checks install.rdf for compatibility. Then it looks for an update.rdf on the web to see if there is a newer version available, or if not, if the current version is compatible with your browser. This would mean that you don't have to repackage your extension until it's time for a new version! This is why I think UMO should have an editable max version.

One of the things that drives most people nuts is having to go through a lot of extra steps to get something simple accomplished. Every time I upload an extension to UMO, I have to manually set the minVersion and maxVersion for Mozilla 1.x. Please make my life easier by including a Mozilla 1.x section in your install.rdf. Even if your XPI only supports Mozilla 1.x and not Firefox/Thunderbird, an install.rdf won't hurt. There is even a GUID for Mozilla 1.x. Here is what I think the extension upload process should look like. Please let me know if you see any process gaps.

• Has install.rdf: Check for existing GUID match
  • New GUID
    • Ask whether theme or extension, etc (could be guessed from dir structure)
  • Existing GUID
    • Check whether logged in user has permission for that GUID
    • Check whether that version is already uploaded
• No install.rdf
  • Have user pick from dropdown of New Theme, New Extension, and updates to which they have permission (best guess from install.js and chrome.rdf?)

If you didn't know, I'm a big fan of Sneakemail. I use it as a pre-filter to direct email to my various accounts, categorize it, and tell me where someone got my address from. Some of my sneakemail addresses are heavily filtered. If you're not coming from specific domains, for example, email sent to my bugzilla account will bounce. Up until yesterday, my email address was certainly not memorable. I finally made use of a facility at sneakemail to give myself one that not only will I remember, but will indicate when it's me that made a comment on a bug. So now you'll see bugspam from Bugzilla-alanjstrBugs instead of 9quawbieby0001. I did run into a small bump in the conversion process, but justdave was able to straighten me out.