The first image below is in Opera, and the second is in Firefox. I wanted it to look like it does in Firefox.

The HTML for the above is:

1
2
3
4
5
6
7
8
9
10
11
12

<table width="550px" border="1">
	<tr>
		<td align="center">
			<table border="1" width="80%">
				<tr>
					<td>Apple Sauce</td>
					<td>Pecan Pumpkin Butter</td>
				</tr>
			</table>
		</td>
	</tr>
</table>

The problem is that the

<td align="center">

is inherited by the tables within that table for IE and Opera but not for Firefox. It’s an easy fix, just make the next <td> into <td align=”left”> to fix it. It’s nothing major, probably very well known…but the first time I’ve come across it.

There was one problem with MyPunchBowl, you were unable to add multiple hosts to an event. Alex emailed support@mypunchbowl.com asking for a solution. That’s the last I heard about it till now.

I happen to use Bloglines, and I have a search in Bloglines for any posts on “Well.ca”. I was a bit surprised tonight when I came across a RSS entry that contained Alex’s email to support@mypunchbowl.com and the subsequent responses. It appears that MyPunchBowl uses Sproutit Mailroom for their support emails, which has an RSS feed of support emails. Someone has subscribed to this RSS feed in Bloglines and was found by my search for “Well.ca”.

When you add a feed to Bloglines, you can mark the feed as “Private” or “Public”. The interesting thing is that marking a feed as “Private” means (from Bloglines): “Private subscriptions don’t show up in blogrolls and you will not be listed as a public subscriber. However, the feed and all its posts will remain available to the public via Bloglines and Ask.com Blog & Feed Search”.

I don’t like putting stuff I want to keep private into RSS feeds and this confirms my feelings.

As a side note, the RSS feed in question can be accessed directly if you have the URL. Maybe I’m being paranoid, but I don’t think this information should be so open.

Django comes with a caching middleware that caches any request made to the site that doesn’t have a POST or GET. You can set the cache to only cache pages for anonymous (using CACHE_MIDDLEWARE_ANONYMOUS_ONLY). The second piece of the puzzle is that the login view for Django sets a test cookie when a visitor first comes to the page, if this is not there, the login will fail.

What happens is:

1. Make a request to the login page without any GET or POST parameters.
2. Web server returns the cached page. (Note: this doesn’t set the cookie because no Python code has been run, the page returned is static)
3. The user enters in the login information and submits the information. This is sent to the server as a POST.
4. The server, because of the POST, now runs the login view. This fails because there was no cookie previously set, but it does set the cookie that should have been previously set.
5. The error message shown to the user is as if it was a failed attempt. If the user reenters the information, the login will now work as the cookie is now set.

Hopefully that makes some sense.

What is the solution? Unfortunately, the caching middleware doesn’t allow you to fine-grain what it caches, it’s an all or nothing. There are two solutions:

1. Always pass a GET param to the login page. By passing a GET param, the caching middleware will not return the cached page.
2. Turn off the caching middleware and instead use the more fine-grained per-view cache.

I’m trying the GET parameter. We’ll see how it works.

There is a Java specification currently in draft for REST, it’s called JSR 311. Now the problem with this is that it is still a draft.

There are currently a few implementations of JSR 311:

• Jersey - Part of Glassfish (SUN’s open source JEE server)
• Apache CXF - This was previously XFire, an implementation of SOAP for Java. It has also started to implement JSR 311. I think this implementation isn’t as fully featured as the other implementations.
• RESTeasy - JBoss’s implementation of JSR 311.

There are a few other options for REST and Java. One option that has been around for a while is called Restlet. The creator of Restlet is involved in the writing of JSR 311 but I don’t believe his project implements JSR 311. Update[11-Jun-08]: As commented below by Jerome Louvel and Stephan Koops there is an extension to Restlet to implement JSR 311, more information can be found on the Restlet Wiki.

Currently, I’m using Apache CXF for my project. But I’m hoping to move to Jersey or RESTeasy because they have a more complete implementation.

The problem with Jersey is that it doesn’t play nice with Hibernate and Maven. This is caused by Hibernate requiring cglib which requires ASM 1.5.3, while Jersey requires ASM 3. There are a few solutions described on the web, but they haven’t solved the problem for me. I’m hoping to work on this at some point, once I find a solution (if I do) then I’ll post it.

• Ali updated the map on the frontpage to be more intelligent and fixed some annoying problems with it. Now it looks better and responds better.
• We now have widgets! See below for an example widget, but this code can be placed into any webpage to give information on a startup or investor listed on StartupIndex.ca. You can access the widget for any startup or investor by the “Embed Widget” option on the profile page.
• A new addition that is not visible from the frontend is a new framework for moderation edits to StartupIndex. This will increase our turn around on approving changes. I’m planning on releasing the code for this as a separate Django application in the near future. More on that soon.

IETester was just released, this tool looks to be incredibly handy. Right now it’s in Alpha, but I’m still planning on using it. Previously I used Multiple IE, but I found that to be a little buggy and prone to not work at all at times. I’m planning on giving it a try sometime today, maybe it will solve at least some problems on testing with IE.

A little late, but Google released an API for loading common JS libraries (such as JQuery, Prototype, Mootools, etc.) from their network of CDNs. This will also increase the chances of someone having the common library cached, instead of having to download an individual copy of JQuery for each site they visit. It’s not a new idea, AOL has hosted the Dojo JS library on it’s CDN for the past few years (more details). Still, I’m glad to see it expand into other libraries. I will probably move StartupIndex.ca to load the JQuery library from Google.

Let us know if you see any problems, or have any brilliant ideas for where the site should go.

You can pass it the excludes field to exclude certain fields from being compared. It does not compare AutoField or RelatedField’s.

1
2
3
4
5
6
7
8
9
10

from django.db.models import fields
def get_changes_between_models(model1, model2, excludes = []):
    changes = {}
    for field in model1._meta.fields:
        if not (isinstance(field, (fields.AutoField, fields.related.RelatedField)) 
                or field.name in excludes):
            if field.value_from_object(model1) != field.value_from_object(model2):
                changes[field.verbose_name] = (field.value_from_object(model1),
                                                   field.value_from_object(model2))
    return changes

Based off of code from Django Modelhistory.

The solution? Force the queryset to a list. For example:

1
2
3

queryset = Widget.objects.all()
wlist = list(queryset)
cache.set("allwidgets", wlist, 3600)

This does mean that you can’t use the cached list as a queryset, so you’ll have to cache it after all the filtering, ordering, etc. has been done.