This has been discussed in a few places and the consensus that I have gathered is that most people would like to have a more conservative definition for purposes of software detection.
Could this be enacted before we hit our "50%" milestone of 50,000 "articles"?
Hitting that number could very well get us undeserved media attention and subsequent media articles that are critical of our article count and average article quality (based on byte size). That is not the type of media attention we need.
For starters, a 500-1000 byte minimum page size should be part of the current spec (keep this in mind though: there are many minor topics that can be considered to be minimal articles with only 500 bytes). Other more complicated criteria can be added later.
Having the default behavior of "Random pages" to be "Random article" would be nice to (that is, only pages that are automatically detected as articles would be displayed).
-- Daniel Mayer (aka mav)