Omni recover review
But for a 1+ MB download size reduction, it might be worth the work. If we were to change omni.ja to !zlib, we'd need to change the code for loading it (libjar), the code producing it (in the build system), and possibly code in the updater (wherever that lives).
#Omni recover review zip file
omni.ja just so happens to be a mostly compliant zip file today. Glandium noted that there's nothing requiring us to use zlib in omni.ja. And brotli's performance across the board (ratio, compression speed, and decompression speed) is often very similar to zstandard and I suspect that the promising results I achieved with zstandard could be replicated approximately with brotli. we should totally do that yesterday." He encouraged me to file a bug, which I'm doing.Īs I noted in the blog post and I'll note here, we don't currently ship zstandard in Gecko. This came up in #developers a few minutes ago and bholley - amused by the 1.5 MB reduction - said "that is huge.
So, if we switched omni.ja from zlib to zstandard, Firefox downloads would be smaller and reading performance may improve due to faster decompress speeds. On top of that, zstandard decompression speed is faster than zlib. So, using zstandard with dictionary compression yields a total compressed size over 1.5 MB smaller than zlib. The total size of an omni.ja with various compression settings are as follows:
I wrote a very long blog post on the Zstandard compression algorithm a few weeks back: īuried in that post are some promising results from compressing omni.ja with zstandard instead of zlib.