Temboz binds to all IP addresses by default today. Add an optional parameter param.bind_address similar to param.host so it can bind to a specific IP address if required.

Suggestion by {link: http://www.tsampa.org/ Kristoffer Lindqvist}

The new CVS version of feedparser dropped the category element and replaced it with 'tags' that has different semantics. This breaks filtering rules that depended on the old semantic. normalize.py should supply the old semantics for backwards compatibility.

The tarball should have param.py.dist and transform.py.dist instead, and Temboz itself should be modified to copy the .dist files over if there are no site customizations.

http://www.gryffyn.com/cgi-bin/rss/DigitalOutbackPhoto_News.xml

The initial deployment of release 0.7 was missing degunk.py because it hadn't been commmitted into CVS. We need an automated acceptance test to verify tarballs in "make dist".

There is a bug in feedparser that causes
to be replaced by semi-random combinations of
>
or
.
. This is particularly visible with Weblogs Inc. blogs. Best to fix this in feedparser.py and contribute the fix back, rather than just a work-around.

When duplicate checking is enabled on a feed, we should check all unread articles for duplicates straight away and remove them.

Make sure Temboz is compatible with the latest versions of SQLite and pysqlite.

This requires a version of pysqlite that can simultaneously access a 2.x or 3.x database (SQLite itself happily lets 2.x and 3.x coexist on the same system).

Newer versions of feedparser support an element "tags". A filtering function tag_any(|_lc|_words) similar to title_any_words would be useful, if only to get rid of Tim Bray's boring flower photos.

Some feeds require HTTP authentication.

The feed at http://www.worldwidewords.org/rss/newsletter.xml only ever has one entry inserted into the database, despite having more than that in the source. It appears to be valid RSS 2.0 according to the feed validator, and feedparser.parse('http://www.worldwidewords.org/rss/newsletter.xml') correctly returns all of the items.

I tried upgrading to 0.8 today and I'm getting the following error message. (I upgraded my database to sqlite3 as you noted).

./temboz --server
46912504432336 1141586308.34 ("select sql from sqlite_master where name='v_feeds'",)
46912504432336 1141586308.37 done
Traceback (most recent call last):
File "./temboz", line 69, in ?
server.run()
File "/home/bdp/Downloads/Temboz/temboz-0.8/server.py", line 376, in run
from singleton import db
File "/home/bdp/Downloads/Temboz/temboz-0.8/singleton.py", line 203, in ?
sql = c.fetchone()[0]
TypeError: unsubscriptable object

This also happens when I try to run with a fresh db.

Traceback (most recent call last):
File "/home/majid/temboz/server.py", line 370, in process_request
self.use_template(tmpl, [self.input])
File "/home/majid/temboz/server.py", line 299, in use_template
tmpl.respond(trans=self)
File "pages/feed_info.py", line 154, in respond
if VFFSL(SL,"getVar",False)('feed_desc', None) and VFFSL(SL,"getVar",False)('feed_desc', None) != feed_desc:
UnicodeDecodeError: 'ascii' codec can't decode byte 0xc0 in position 0: ordinal not in range(128)

for people interested in the bleeding edge

The sort by SNR option in the AllFeedsPage and the default sort order for that page (feeds with unread articles first, sorted by SNR, then feeds without unread articles, sorted by SNR) are broken when Temboz is running SQLite 3.x/PySQLite 2.x

Feeds sometimes move and the old feed redirects to the new one. If the redirection is a permanent one (HTTP status code 301), Temboz should save the new URL as the XML URL for the feed.

Often when this happens, the permalinks/guids in the feed change as well (even though they shouldn't), leading to duplicate posts. An optimization would be to filter out items with duplicate titles after the feed has changed locations.

We should probably create a pseudo-article as a service notification so the user is aware something changed with the feed, along with a list of duplicate items suppressed.

There is a potential privacy issue with this: a feed owner might redirect the user to a feed URL with some tracking code embedded in it for tracking purposes. Putting the URLs in the feed URL change service notification would allow the user to be aware of this if it happens.

Hey, temboz looks great. However, I'd like the ability to run it as a cgi-script (or mod_python) like alot of typical server side aggregators, as I use external hosting and am not enabled to run daemons on that machine.

Thanks

Autodiscovery is not working for http://www.lifehacker.com/
Nor for http://www.laughingsquid.com/

When a feed exceeds the max_errors threshold in param.py it is automatically suspended. Automatically generate a service notification when this happens instead of doing it silently.

Temboz is hanging during an update. The server will accept connections then hang. The log shows the following error message:

Traceback (most recent call last):
  File "/usr/local/lib/python2.4/threading.py", line 442, in __bootstrap
    self.run()
  File "/home/majid/temboz/update.py", line 476, in run
    update()
  File "/home/majid/temboz/update.py", line 463, in update
    update_feed(db, c, *feed_info)
  File "/home/majid/temboz/update.py", line 261, in update_feed
    process_parsed_feed(f, c, feed_uid, feed_dupcheck)
  File "/home/majid/temboz/update.py", line 378, in process_parsed_feed
    c.execute(sql)
  File "/home/majid/temboz/singleton.py", line 16, in execute
    result = self.c.execute(*args, **kwargs)
  File "/usr/local/lib/python2.4/site-packages/sqlite/main.py", line 244, in exe
cute
    self.rs = self.con.db.execute(SQL)
TypeError: execute() argument 1 must be string without null bytes, not str

Exception exceptions.AssertionError: <exceptions.AssertionError instance at 0x85
7f58c> in  ignored

When Temboz is run on SQLite 3.x/PySQLite 2.x, sometimes when you do an action that updates the database like clicking on "Thumbs down" while the hourly fetch is done, Temboz will deadlock, and has to be killed.

It would be good to have a report that shows trends in the SignalNoiseRatio, e.g. calculated over the last month, last three months, entire history, like a stock chart, to identify feeds that are losing their relevance and are thus possible candidates for culling.

When you update a filtering rule, you go back to the top of the rules list. It would be better if the list scrolled down to the rule that was updated.

Another option would be to update rules using AJAX.

Currently, there is no timestamp set for when an article is flagged as interesting. This means there is no way to sort interesting articles in the order they were read, as opposed to the order they were received by Temboz.

SQLite's handling of concurrency is not very good. Things crawl to a halt when trying to read and write at the same time, e.g. when trying to read feeds as the update thread kicks in. A database with multi-version concurrency control like PostgreSQL would do a much better job.

Users should be able to enter annotations on an item, so they can share an annotated list of interesting links rather than one without comments as today.

I updated my Temboz install to the CVS version (CVS-2006-03-06) to try the new buttons, but now it appears that it is not updating my feeds. I can force an update with --refresh, but --server mode doesn't seem to be doing update polling automatically. My interval in param.py is set at the install default of 3600.

I poked around a wee bit, and tried turning on the Debug mode in param.py but I couldn't figure out a workaround. I tried running CVS-2006-03-10 but it didn't seem to make a difference. I have not tries switching back to 0.8 yet. I'm wondering if I missed something important when changing versions. I just copied my 0.8 db file over.

Peter Janes reports he uses "Thumbs up" to sideline articles for future consumption, rather than interesting/uninteresting articles. This is a perfectly common use case that should be fully supported.

This is 2005, the "subscribe to a feed" functionality needs autodiscovery.

I am lazy and use the same password for most non-sensitive sites (and another password for sensitive sites like online banking). This is clearly insecure and subject to catastrophic failure.

Temboz is as good a place as any for me to implement a single-sign-on system that would generate a domain-specific password for a site, that way if my password is compromised, the damage is contained within that site.

Of course, this feature has nothing to do with feed reading...

http://www.russellbeattie.com/notebook/1008751.html

Sometimes a filtering rule is over-zealous. It would help debug such rules to link the article to the rule that caused it to be filtered, either in the log, or better yet in the fm_articles table itself.

This is not perfect, as the filtering rule text may have changed between the moment it was evaluated to filter out the article and the moment the user checks it, but it's better than nothing.

Packaging Temboz for easier installation using Py2app or Py2exe with all dependencies included would make it easier to use for non-Pythonistas.

When you Command-click or middle-click on an article title in Safari, it gets highlighted in yellow as if you had clicked on it or clicked on the thumbs-up icon.

The SignalNoiseRatio report and sorting do not currently take time into consideration. This is problematic, as feeds can lose relevance over time. Implementing something like an exponentially decaying SNR (like that used to compute UNIX load averages) would improve the usefulness of this feature without unduly penalizing high-quality, ultra-low-volume feeds the way an arbitrary cut-off to the last N days would.

Photos exported from Flickr to Typepad blogs have embedded <style> tags in them. Feedparser strips them out for security reasons (although they could probably be sanitized the same way style attributes are). However the tags themselves are stripped but not the contents, so the CSS code appears in the entries instead of being filtered or sanitized.

{image: screenshot.gif}

Filtering rules should have a title field so we can give a complex rule an easier to scan for label like "annoying memes".

The main view can switch between showing unread, interesting, uninteresting or filtered articles by respectively adding ?show=unread, ?show=up, ?show=down or ?show=filtered to the URL, but there should be a way of doing this from the UI.

Peter Janes reports the feedparser debug output in the FeedDetailsPage is sometimes not escaped correctly and shows images, etc.

This is probably due to feeds with a literal in the feed text. Any occurrences of such should be escaped to prevent a potential security risk, not to mention UI breakage.

I recently installed Temboz. I like it so far, but one thing I noticed was that I end up scrolling around a lot. To be more specific, I read an article and then I have to scroll back up to click on the Interesting or Uninteresting buttons.

It would be nice to either have those display at the bottom and the top or to have an option of having them display at the bottom.

http://www.kbcafe.com/rss/usm.html

There is a potential race condition when editing rules - you might have modified the rule you are editing in another window, and thus unwittingly edit a stale entry. The rules submission should use something like a MD5 hash of the old rule content so that attempts to submit an edited version of a stale rule can be detected and the conflict resolved by the user.

The OPML import module depends on the {link:http://www.reportlab.org/pyrxp.html PyRXP} XML parser module, which is not very commonly found package. Using on of the alternative XML parsers shipped in the Python standard library would make installation easier for users.

One CreativeUseForTemboz is as a filter for feeds like Craigslist apartment search or Techbargains.com. When time is of the essence, sending an email notification of new articles could make the difference between getting the deal or not.

The feedparser dump in the feed details page should really be moved to a page of its own. It is wasteful to fetch a feed everytime a user want to check on feed status.

Feeds are polled hourly by Temboz. It does use HTTP mechanisms like Etags or If-Modified-Since to avoid downloading an entire feed if it hasn't changed (not sure about mod_gzip support, though), but for many feeds a daily or even weekly update would be just fine.

One way would be to use statistics like item inter-arrival times, but we probably should have a user control that sets hourly/daily/weekly/monthly/automatic updates.

Adding del.icio.us/Flickr/Technorati style tagging to interesting articles would make them easier to find after the fact.

It would be very valuable to have articles in the MainView be sorted by SignalNoiseRatio, so articles from high-quality feeds stand out from the rest when you are in a hurry.

Possibly, instead of a simple priority sort based on SNR, do something like a weighed sort like order by (current_date - date_created)/snr

Some sites have multiple feeds listed in autodiscovery: main feed, feed with excepts, comments, and so on. Temboz should offer a way to let the user select between them.

For simplicity's sake, when both ATOM and RSS are offered, don't consider these as different variants, only when you have multiple ATOM or multiple RSS feeds listed.

FilteringRules are powerful, but difficult to manage if there are too many. It would make sense to reduce clutter on the filtering rules page by moving feed-specific rules to the FeedDetailsPage instead.

Temboz purges uninteresting articles after 6 months. Normally, this is after 6 months or the oldest article in a feed, but for slowly updated feeds (including my own), the detection of which article is oldest is not working correctly as purged articles reappear.