http://www.wired.com/wiredenterprise/2012/12/github-bots/
gitbots...
So you have a technique that makes lots of people code better (e.g. a better compression alg for your images). You create a bot that looks in github repos, find projects that have images, checks wether the new compression algorithm can make a difference , and if it can you automatically do a fork/recompress image/submit pull request back.
While the negatives are quite obvious the positives are more interesting.
Improving source code at scale is really interesting - and feasible because of open source collab platforms like github.
You can have a someone run a spell checker against doc/readme pages/ and when it find mistakes pass them through some crowdsourcing and then automatically fix everyones docs. You may find that people haven't done man pages out of their wiki - creating man pages and automatically add the man page install in the install process could be automated. People may not have hooked up a continuous integration system like travis that confirms on commit...even though they have the tests.. thats again automatable to get the travis-ci treatment.
If people are annoyed by the pull request noise.. you can have the bot submit an issue asking the repo owner if they are interested for the fix and only proceed if the issue doesn't close/gets a positive reply.
If people are annoyed by the issue noise from bots... they should have a bot that filters/screens issues based on user that the bot is associated with or any other rule they need.
gitbots...
So you have a technique that makes lots of people code better (e.g. a better compression alg for your images). You create a bot that looks in github repos, find projects that have images, checks wether the new compression algorithm can make a difference , and if it can you automatically do a fork/recompress image/submit pull request back.
While the negatives are quite obvious the positives are more interesting.
Improving source code at scale is really interesting - and feasible because of open source collab platforms like github.
You can have a someone run a spell checker against doc/readme pages/ and when it find mistakes pass them through some crowdsourcing and then automatically fix everyones docs. You may find that people haven't done man pages out of their wiki - creating man pages and automatically add the man page install in the install process could be automated. People may not have hooked up a continuous integration system like travis that confirms on commit...even though they have the tests.. thats again automatable to get the travis-ci treatment.
If people are annoyed by the pull request noise.. you can have the bot submit an issue asking the repo owner if they are interested for the fix and only proceed if the issue doesn't close/gets a positive reply.
If people are annoyed by the issue noise from bots... they should have a bot that filters/screens issues based on user that the bot is associated with or any other rule they need.
No comments:
Post a Comment