529173 - Replace search.json with startup cache

Status: RESOLVED WONTFIX

(Firefox :: Search, defect)

Comment 1

[Richard Newman [:rnewman]]

Taras: I assume that this bug is basically "don't dump 160KB of JSON to disk; let's store something precompiled", right?

Is there a reason for startup cache instead of BSON, or even writing favicons separately?

We want something like this to avoid chewing up so much space in Fennec profiles.

Comment 2

[(dormant account)]

Richard, this was more to do with data locality. less files to read => faster. As far as formats, bson sucks. json + gzip or json + lzma win over every binary json format + compression I tried.

Startup cache is nice because it's a zip, so you get compression for free.

Comment 3

[Richard Newman [:rnewman]]

Great. So long as we get a space saving, I'm happy :)

What's the next step for getting this moving?

Comment 4

[(dormant account)]

(In reply to Richard Newman [:rnewman] from comment #3)
> Great. So long as we get a space saving, I'm happy :)
> 
> What's the next step for getting this moving?

implement it?

Comment 5

[Richard Newman [:rnewman]]

Heh.

Let's see if these fancy new backlog flags make a difference…

Comment 6

[:Gavin Sharp [email: [e-mail]]]

I don't know what the breakdown of profile space on Fennec looks like or how many thousands of papercuts you're suffering from, but 160KB doesn't seem very big to me, so without context the size argument isn't particularly convincing.

This seems like it would be a significant chunk of work, and there's little evidence that it will help startup significantly (particularly now that search initialization is all async), so I don't think there's enough win here to justify the cost.

Comment 7

[Florian Quèze [:florian] (unavailable until May 22)]

I'm likely to replace search.json in bug 1203167. If the size of the base64 encoded icon files is a problem, the bug to fix it is bug 986676.

Comment 8

[Mark Banner (:standard8)]

search.json is now a compressed lz4 file. Not sure if the icons are still an issue or not - but if so, that's bug 986676 as mentioned in the previous comment.

As this doesn't seem to be a current issue with startup, I'm going to close this out. We can always re-open or file a new issue if we find something significant.