Compression of dumps

[xendk]

SQL dumps tends to be large without compression. We should compress them per default.

[kasperg]

Does this require much more than adding --gzip when running sql-dump?

[xendk]

That would compress it, yes, but then omg would break: https://github.com/reload/deployotron/blob/master/deployotron.actions.inc#L826. Secondly it would make #10 more difficult (tailing an uncompressed file is so much more effective).

One way of fixing this and working #10 into it without wasting the diskspace would be to drop --result-file and pipe the dump though awk or sed with a small script that simply prints a magic message to stderr when seeing the "dump completed" line and piping it's output directly to gzip. Then #10 would be fixed by checking that the command outputted the proper magic message, and we'd have a compressed dump.

[xendk]

Actullly, after careful reading of both master and 6.x branch of Drushs implementation of sql-dump, it turns out that run gzip as an extra command on the file afterwards. They only pipe directly to gzip when not using result-file.

So basically, doing the dump, checking the file for completeness and then gzipping isn't worse off that what it's currently doing.