I wrote a crawler in 2005 in C.  I meant for it to be the fastest, least-nice
crawler that I could build.  I've modified it several times over the years and
wrote some wrapper scripts for it to help with the de-duping and url-management.

This is not a well-packaged tool that's plug-and-play.  It's a collection of a
bunch of scripts and tools that I've written to do different jobs for different
projects.

I used this crawler to generate the URL seed list for archiveteam.org for the
geocities.com rescue project.

Now, I'm releasing it into the public via github so that other people can use it
and make good use of it.

Features:

* Super/mega fast and efficient.  100 threads on linux should be do-able on the
weakest of machines.
* Doesn't care about robots.txt.  If you want a nice crawler, there are hundreds
of them out there.  This is for ripping a site as fast as possible.

Requires:

* libxml2-devel
* libcurl-devel
* pthreads
* pcre-devel

Notes:

* fakesite/crawlme.php is a php script that simulates a site that you can use to
test the crawler.

  example: .../crawlme.php?numurls=100&depth=4 will provide a good bunch of urls
to crawl with the crawler.

* crawl.dat is the file of urls that the crawler will grab.
* output/timestamp is the output directory for each crawl.
* I'm working on a URL-distribution server for running these crawlers on multiple
machines but I may start that as a seperate project.

If you want to contribute anything to the project, feel free, however, my general
philosophy about the crawler is to keep it as lean and mean as possible.  No
SQL, no fancy stuff.  Keep using the unix tools (text utils, pipes, threadding)
and stay away from bloated technologies (ruby, sql, ...) and I'll be really happy
to add your contributions to the project.

TODO:

* perhaps put the BerkeleyDB management into the crawler, so it's more self-
contained.
* add queueing to each thread, so each thread pulls more than one url to work on.

- Steve Webb ([email protected])