Archivebox: Archive Method: Add WARC file output

[bmix]

An overview of existing projects to consume them is here:
http://www.archiveteam.org/index.php?title=The_WARC_Ecosystem#WARC_viewer

[pirate]

See: https://github.com/pirate/pocket-archive-stream/issues/11

[eqyiel]

Can this be re-opened? The linked issue is about the webarchive format, which is a totally different thing.

https://en.wikipedia.org/wiki/Webarchive
https://en.wikipedia.org/wiki/Web_ARChive

WARC is the one that archiveteam and libraries have thrown their weight behind: http://fileformats.archiveteam.org/wiki/WARC

[pirate]

Oh I didn't know they were different, thanks @eqyiel.

[pirate]

Just requires adding a new config FETCH_WARC option and archive_method.fetch_warc:

https://www.archiveteam.org/index.php/Wget_with_WARC_output

[anarcat]

for the record, pywb has trouble reading wget WARC file as its output is non-standard: https://github.com/webrecorder/pywb/issues/294 you might want to consider another crawler for the task or see that wget fixes their stuff first.

[f0086]

The bookmark-archiver is mentioned in a recent LWN article:
https://lwn.net/Articles/766374/

WARC archives seems the be the way to go for archiving webpages, especially for SPAs and other JS-heavy and streaming pages. Wget is -- according to that article -- a little broken in terms of WARC output, but there are a list of alternatives which works fine.

It would be very nice if bookmark-archiver get support for WARC archives.

[pirate]

I saw the article, and I actually emailed @anarcat about it since it turns out we both live in Montreal.

I think if I add WARC saving I want to do it with https://github.com/internetarchive/brozzler and a headless browser to cover all the edge cases around JS execution.

[FiloSottile]

The easiest way to do this might be to just point the headless browser you already run to a WARC recording proxy. However brozzler looks like a nicely maintained piece of software (and way better than the old Java stuff the IA was using before), so maybe migrating to that can be a longer term goal.

EDIT: just noticed there's #113 for brozzler already.

[pirate]

Wget WARC file output is now supported in e8808b0, via the FETCH_WARC=True flag (on by default).

I'll investigate creating another a more complete warc including dynamic requests with the chrome proxy, but for now the simpler wget one is worth testing in the interim.

[FiloSottile]

WARCs are nice for bundling all resources needed to render a page in a browser, but using wget generates a WARC which is just a container for the HTML. Maybe open a new issue for the browser integration?

[pirate]

Yeah, actually I think I'll turn it off by default for now because it's too minimal to be a default until we get warcprox working with headless.

Here's the new issue to track the all-in-one WARC file https://github.com/pirate/ArchiveBox/issues/130