Today's Work

I started the day by trying to learn Microsoft Project. I found a short tutorial on the web and studied it. Then I started modifying alias.pl to convert it to a more robust link checker. The current version does not specify which page bad links are found on and there is no way to add additional functionality such as html checking to the script. I mean to address both of these issues by rewriting the script in an object oriented style. I mdified the standard html footer to include the current user as well as the curator of the html page. I wrote a python script called "curator" (naturally) to get this second piece of information. It searches the local roles in the current folder and all parent folders to get this information.

Today's Work

Today I tested the changes I made yesterday to STMethod and they worked. So now methods throw an error when they are displayed as the entire page, but not when they are displayed as part of a page. I asked Matt Kromer about using the existence of the RESPONSE object to determine whether the method is whole or part, and he wasn't too keen on it. But for lack of anything better, we'll keep it.

Then I went for broke and modified STObjectBase to put up a login form instead of throwing an error when notViewable is called. That change worked also and I finished up by taking the audience testing code out of isViewable and making a new method, myAudience. I committed the changes to CVS and wrote up a test plan. I got Jim to test my changes and in return I tested his change to fix the base tag problem. So both changes are ready to go to production.

After that I wrestled with Microsoft Project and made some changes, but didn't finish getting the Remedy project updated. I also looked at alias.pl to see what changes will be necessary to change it into a proper web link changer, since that is all it is being used for. But being late on a Friday afternoon, I decided it was best not to start changing it until Monday.

Today's Work

I talked with Greg about the mechanism to notify people when a 911 call is received. It seems the mechanism is in place but the list of people to notify is not populated. I will get this list from Glen.

The webinator services catalog was too small. I changed the permissions on the account which builds the catalog and that seemed to resolve the problem. There still are a few sort_on links that need to be removed from the catalog. And there are issues with multiple urls pointing to the same page. But we can go with what we have now.

The issue with the links that don't do anything when you click them remains to be resolved. The problem is that unauthorized methods return an empty string (by design). I am trying to determine if the method is called as the entire page or part of a page and throw an error if the former is true. The folks at Zope say this can't be done, but I am testing some code to try it. The jury is still out on whether it works. Ideally we would like to put up a login form instead of throwing an error, but I have yet to figure out a way to do this.

Today's Work

I did a small amount of work on the remedy project in the morning I need to finish updating the wiki and the project plan.

I finished a script that analyzes the ARS log file and prints out the 20 longest running command sequences. I ran it on a log file and the longest running command (813 seconds) was a full text search, followed by a lot of escalation commands.

I spent a lot of time looking at our problems with web searches. It seems that our problems relate to how authentication works or doesn't work in Zope. A lot of ideas were talked about, but there was no definite result.

Greg found a javascript example site and Leigh suggested a bookmarklet site.

Today's Work

I looked at the output from the center/services search some more. The output had the problems previously noted: links to edit pages and links that wouldn't do anything when you clicked them. I suggested some changes to the script that builds the search catalog to Mike. The fixes were not correct and the resulting catalog only has a single record. I wrote a short perl script that prints out the http response header to see what was happening with the links that don't do anything. The pages return a 204 error (no data on page). It's not clear whether this is a permissions problem or not.

At Greg's request I started writing a script to parse the log file that ARS produces. The object is to identify which queries are taking the longest time to process.

Today's Work

This morning I looked at the center/services search some more. I figured out how the abstract is built (it's a built in Webinator function). I concluded that there are no real problems on the search end, the problem is in how the Webinator database is built. After lunch there were some problems in moving my STWebSearch change to test. The solution turned out to be adding the -A flag to the cvs update command, which forces it to get the latest version. Part of the problem with the Webinator results looks to be caused by aliasing: two or more urls pointing to the same page. I also talked with Greg a bit about Glenn would like to see the reports modified.

Today's Work

I started the day out by putting the test plan to move the web search changes on the wiki. Then I discussed the reports Greg had generated with Glenn and put together a preliminary project plan and sent it to Rick Singer. Then I looked at the Webinator documentation and found nothing connected with our searching problem. But I guessed the style sheet was showing in the Webinator database because it wasn't hidden by comments. My guess was right, and I asked Mike Watterson to move the change to test. I still need to understand how the program is producing the abstract of the page that is put in the search results.

Mike Wiggs tell me that sitecheck runs extremely slowly. I'll have to check it out.

The Daily Python give links to current articles of interest to Python developers.

Today's Work

I tested the Center search page, which works, removed the debugging code from STWebSearch, and committed the changes to CVS. I gave Mike a list of web pages in the Center site that should not be indexed. I helped Calvin debug a problem that prevented his CGI script from creating an html file.

I'm having problems with the mouse on my computer, which is disheartening, because I had exactly the same problem with my previous mouse.

Today's Work

Today I resolved most of the issues with the websearch. It now accesses the correct database and correctly switches between the services database and the stsci database. Remaining issues are setting up the center database (if any) and taking out the diagnostic writes.

Today's Work

The day was spent working on the web search. I got the web search to use the correct database by modifying the Vortex script. But we found that the databse had screwy content. So Mike rebuilt it and now I can't connect to it again. I modified the Python product that accesses the database so it can vary the scope and path variables simultaneously. This change also needs to be debugged.