2003 08 27

Base | Research


A better Monster

Monsterboard is a great site, because it really has tons of job opportunities listed. Their business model ensures that the signal to noise ratio stays very low. (putting noise on Monsterboard is a very expensive hobby).

That said, I find that its usability lacks some features. Specifically, I find that I do searches, also do the related searches and then spent an inordinate amount of time looking at the same job ads, because it is impossible to automatically detect that I have already seen (an in most cases rejected) the ad. Especially when doing broad searches, going back over a good two months, this gets tedious pretty quickly, fast ADSL notwithstanding.

The model I am looking for is rather simple: do one or more searches, uniquefy the result set, filter it through a list of already rejected ads and present the remainder for evaluation. If this had to be built into the site, there could arguably be some logistical problems. For one thing, every user would have to have an entry in the database (easy) with associated status for every possible job ad available (not too difficult either, but a lot of work for probably not very much gain). And that is just the datamodel. Implementation, taking into account multiple simultaneous connections, a sizeable data set to work with and a user interface that is not geared towards this is a pretty hairy beast.

Not that I have any other option, but I will shortly have a go at solving this problem client side: do one or more searches, parse the output and repeat the process going forward. It will involve some scraping, but it looks easy enough.

2003 08 26

Base | Research


Monstrous URLs deconstructed

Monsterboard uses a dreadful system of URLs, in which they encode enourmous amounts of information about the path a user has taken to get to a certain page. This relates to the bulk of useful pages on their sites: the ones with job ads on them. An example:


As a rule, I include URLs like this one in emails I send as a response to such a job. (not this one, mind you). I got fed up with having to include an apology for the fact that most likely, the email recipient would have to manually paste the URL back together again. Going through URL shortening services like the ones listed here seems like overkill for this one-time usage. In addition to that, I am reasonably afraid it may confuse some recipients that are not necessarily well versed in modern technology. Thus, I had to try to come up with a better idea.

The fact that you can easily run into the same job ad again and again through different searches, where the colour of the link on the page indicated that my browser thinks it has not visited the page before put me onto the idea that there might be search-dependant information buried in the URL itself. A cut-n-paste later confirmed that. For entertainment value, here is an overview of the parts of the URL that I could identify.


This obviously is the host and base URL.


This looks to me like a unique identifier of a job ad. (yeah, yeah, very few points for figuring that out.)


This looks a lot easier when de-URL-encoded:

AVSDM=2003/09/01 11:24:00

I am guessing this is the date a job ad was posted. Interestingly, this is not something that can be determined from the page itself. For any given working link to a job ad, however, you can usually navigate away from it and then back, for instance through 'click here to see all [blah] opportunities', then back to the one you came from. The URL in your address bar will have changed and will now at least contain this information.


It seems obvious that it has something to do with a logo, but I have yet to find out which one. Leaving it out of the URL makes no visible difference to the resulting page.

col=dltci cy=US brd=1

Three items that might be 'column', 'currency' and 'border', but their meaning and effects elude me so far.

lid=157%2C184 lid=157,184

These are locale identifiers. To find the ones that are good for you, do a manual search, pick the first job of the list and grab the numbers. I have found that repeating the lid argument works fine.


An argument I have only ever seen empty.


This should obviously be your query.

Some trial and error showed that to refer to a page with a job ad, really the only thing that you need is the Job ID. Thus, this will suffice:


2003 08 21

Base | Navelstaring


Britt's page is up

Obviously, when just mucking about with a new site, the first thing one does is stick up pictures of one's newborn child. Here is a page about Britt.

Base | Navelstaring


And now for a blog

The main purpose for this site is to hold my (as of yet non-existent) blog. I am thinking or using Blosxom, but I haven't decided yet whether to host dynamically or to batch produce and upload. I am leaning towards the latter, but the deafult operation and hence all the support and tooling is geared towards dynamic serving.

Base | WebDev | CSS


Stylesheet uploaded

I happened to have written not just a ton of documentation for my current employer, but I also did a stylesheet for it. For lack of a quicker alternative, this at least looks somewhat decent.

Base | Navelstaring


Page here but not quite finished...

The page is here, but as you can see it is not quite finished. I really do not approve of big black-and-yellow 'under construction' banners, so a bit of text will have to do in explaining that I am really in the very early stages of setting this thing up.