833098 - Kick RDF out of Firefox and Gecko

Status: RESOLVED FIXED

(Core Graveyard :: RDF, defect)

Description

[Marco Bonardo [:mak]]

Among the various reasons for which rdf is old, unmaintained and so on, it's also another piece of libxul that may go away helping out PGO linker memory usage.

Comment 1

[Marco Bonardo [:mak]]

I couldn't file a bug for install.rdf, is there any plan for moving add-ons out of it?

Comment 2

[Kyle Huey (Exited; not receiving bugmail, old account, do not use)]

Yes please!

Comment 3

[Mike Hommey [:glandium]]

comm-central is also full of stuff using rdf.

Comment 4

[Marco Bonardo [:mak]]

there are also xul templates that likely we may drop...

search.rdf, panels.rdf and downloads.rdf are likely old versions of our dbs, there is lots of code and test codes accessing these that most likely we don't need anymore... reducing globally our rdf usage by removing pre FF3 code would be good, at least to reduce a bit the code base using it.

(In reply to Mike Hommey [:glandium] from comment #3)
> comm-central is also full of stuff using rdf.

yes, though we may disable it in confvars.sh just for Firefox, for example, these kind of projects can be done in parallel

Comment 5

[Axel Hecht]

IIRC, we removed the disable-rdf configure flag because it was rather broken for content internals, or somewhere there.

Comment 6

[Joshua Cranmer [:jcranmer]]

There is another use of RDF that hasn't had a blocking bug filed: the charsets menu.

Comment 7

[Benjamin Smedberg]

See bug 357276, a module I wrote which exposes a JS API for parsing existing RDF data that doesn't require any of the C++ goop (the expat integration may be horribly broken, but the logic should still be pretty good). That should at least provide a simple way of importing existing data like extensions.rdf into whatever new format we pick for that data.

Comment 8

[Dave Townsend [:mossop] (PTO until 6/3)]

(In reply to Marco Bonardo [:mak] from comment #1)
> I couldn't file a bug for install.rdf, is there any plan for moving add-ons
> out of it?

No hard plan, just talk. But we could switch to a JS based parser for install.rdf and update.rdf files.

Comment 9

[Marco Bonardo [:mak]]

(In reply to Dave Townsend (:Mossop) from comment #8)
> (In reply to Marco Bonardo [:mak] from comment #1)
> > I couldn't file a bug for install.rdf, is there any plan for moving add-ons
> > out of it?
> 
> No hard plan, just talk. But we could switch to a JS based parser for
> install.rdf and update.rdf files.

This would be a great idea, also considering comment 7
I hardly believe we need all of the rdf functionality, a "simple" parser can likely cover our needs. Maybe we should discuss this project at the next work week.

Comment 10

[Ben Bucksch (:BenB)]

> rdf is old, unmaintained and so on

Black humor: Quite macabre statement, after Aaron Swartz died about a week ago. :-/
http://en.wikipedia.org/wiki/Aaron_Swartz#W3C

Comment 11

[Ben Bucksch (:BenB)]

(At least you didn't call the bug "Kill RDF"...)

Comment 12

[Marco Bonardo [:mak]]

Not sure what you are trying to achieve by that, my words were clearly pointed at our implementation of the standard and the fact we don't maintain it.

Comment 13

[Marco Bonardo [:mak]]

I suppose you prefer if I talk about "rdf/"

Comment 14

[Axel Hecht]

As we're side-tracking. RDF is maintained, I'm the module owner of it.

Anything that we do with RDF ties us to functionality that's not conforming to specs or unfortunate uses of it. Which means that anything that blocks us from killing RDF is also keeping us from improving out implementation, or merely just add tests for it.

That makes a 100% feature freeze the only option for our current impl.

Given that RDF isn't considered sexy these days, killing it is the better motivation to drop the users of unspecified or badly designed functionality, compared to "I'd like our RDF impl to be spec compliant and test-suite-passing".

Comment 15

[Ben Bucksch (:BenB)]

FYI, comment 10 was merely a joke, nothing else.

Thunderbird depends heavily on RDF, so removing it entirely from the tree would pose significant problems for TB. jcramner is discussing on tb-planning how to do it, but it's quite an effort. Previous attempts have caused serious regressions and thus slowed down.

Comment 16

[Mark Banner (:standard8)]

(In reply to Ben Bucksch (:BenB) from comment #15)
> FYI, comment 10 was merely a joke, nothing else.
> 
> Thunderbird depends heavily on RDF, so removing it entirely from the tree
> would pose significant problems for TB. jcramner is discussing on
> tb-planning how to do it, but it's quite an effort. Previous attempts have
> caused serious regressions and thus slowed down.

As comment 4 has already suggested, a build time option would suffice for now if Firefox has finished core removals before we get that far, and assuming we don't get impacted by the issues in comment 5 (I strongly suspect with the removals of the dependencies on RDF from core, we'd be able to make it work).

Comment 17

[Henri Sivonen (:hsivonen) (on leave)]

(In reply to Joshua Cranmer [:jcranmer] from comment #6)
> There is another use of RDF that hasn't had a blocking bug filed: the
> charsets menu.

Firefox no longer uses RDF for charset stuff. Getting the dead code out of m-c is bug 943252.

Comment 18

[Kris Maglione [:kmag]]

Attachment: Bug 833098: Part 1 - Remove dead code in xpfe directory viewer.

Most of this module is dead code, intended to convert directory listings into
RDF data sources for a XUL front-end which no longer exists.

Rather than keep the remaining code which still has any effect, I opted to
just replace it with a JS stub, which winds up being much simpler.

Review commit: https://reviewboard.mozilla.org/r/240606/diff/#index_header
See other reviews: https://reviewboard.mozilla.org/r/240606/

Comment 19

[Kris Maglione [:kmag]]

Attachment: Bug 833098: Part 2 - Remove RDF service.

Review commit: https://reviewboard.mozilla.org/r/240608/diff/#index_header
See other reviews: https://reviewboard.mozilla.org/r/240608/

Comment 20

[Kris Maglione [:kmag]]

Jorg, it looks like Thunderbird still uses the RDF service pretty extensively. Presumably you'll want to move it to comm-central.

Comment 21

[Jorg K (CEST = GMT+2)]

We have been looking at removing RDF for a long time, see meta-bug 420506. Since we're chronically understaffed, we're always in fire-fighting mode. When XUL templates were removed, we landed bug 464710 (there was an RDF removal component in there), now we have bug 453908 and bug 457333 in the pipeline.

What's the ETA for landing this? Maybe it could wait for mozilla62 while we're still busy with TB 60 preparing the ESR.

Comment 22

[Kris Maglione [:kmag]]

I was aiming for the start of the next cycle. This is a lot of dead code to keep shipping after nothing in-tree uses it, especially given the number of relocations it requires in the content process.

If comm-central still needs it, it should be pretty trivial to move there.

Comment 23

[Henri Sivonen (:hsivonen) (on leave)]

Comment on attachment 8971850 [details]
Bug 833098: Part 2 - Remove RDF service.

https://reviewboard.mozilla.org/r/240608/#review246318

Looks good to me. Thank you! It's unclear to me, however, why review is requested from me. I'm not the module owner for this code. Axel Hecht is.

Comment 24

[Kris Maglione [:kmag]]

(In reply to Henri Sivonen (:hsivonen) from comment #23)
> Comment on attachment 8971850 [details]
> Bug 833098: Part 2 - Remove RDF service.
> 
> https://reviewboard.mozilla.org/r/240608/#review246318
> 
> Looks good to me. Thank you! It's unclear to me, however, why review is
> requested from me. I'm not the module owner for this code. Axel Hecht is.

Ah, I was mostly asking for review for the DOM and toolkit changes. I thought RDF was unowned at this point.

Comment 25

[Axel Hecht]

Moving this over to Core RDF.

I've glanced at the patch and I don't see anything concerning on the technical and Firefox level. Before I flag an r+, though ...

(In reply to Kris Maglione [:kmag] (long backlog; ping on IRC if you're blocked) from comment #22)
> 
> If comm-central still needs it, it should be pretty trivial to move there.

I'd have to disagree with that assessment. As the current module owner, I expect my duty to end with the removal of the code in mozilla-central.

I've pretty much succeeded in keeping the rdf module from changing. Yet, it's one of our oldest piece of C++ code, and as such is frequently hit by refactorings in the underpinnings there, https://hg.mozilla.org/mozilla-central/log?rev=rdf%2F.

There's also a bug component with a number of bugs that would move to the graveyard if the code was just removed. Otherwise we'd have to move the component, maybe to MailNews Core? Not sure if Emma would insist on a Triage Owner there.

While it might be trivial to copy the existing code over, it's non-trivial to keep it compiling, and there's a couple of module-ownership changes that we'll need to make alongside.

Flinging this through Jorg's needinfo queue once more.

Comment 26

[Jorg K (CEST = GMT+2)]

I'm not sure what you want me to answer here. I'll finally get rid of RDF as per comment #21. We have no intention of moving code to mailnews.

If this can wait until mozilla62, we'll probably be fine.

Comment 27

[Axel Hecht]

Comment on attachment 8971850 [details]
Bug 833098: Part 2 - Remove RDF service.

https://reviewboard.mozilla.org/r/240608/#review246424

Sounds like a plan, thanks Jorg.

Let's get this done with early next cycle.

Drum roll.

Comment 28

[Jorg K (CEST = GMT+2)]

(In reply to Kris Maglione [:kmag] (long backlog; ping on IRC if you're blocked) from comment #20)
> Jorg, it looks like Thunderbird still uses the RDF service pretty
> extensively. Presumably you'll want to move it to comm-central.
After looking at the main open bugs we have for removing use of RDF (bug 453908 and bug 457333 and bug 732106), I think I'll have to change my mind and fork RDF at least for a few months :-(

Comment 29

[Dave Townsend [:mossop] (PTO until 6/3)]

Comment on attachment 8971849 [details]
Bug 833098: Part 1 - Remove dead code in xpfe directory viewer.

https://reviewboard.mozilla.org/r/240606/#review248012

Comment 30

[Dave Townsend [:mossop] (PTO until 6/3)]

Comment on attachment 8971850 [details]
Bug 833098: Part 2 - Remove RDF service.

https://reviewboard.mozilla.org/r/240608/#review248014

Comment 31

[Kris Maglione [:kmag]]

https://hg.mozilla.org/integration/mozilla-inbound/rev/2ea9f1f5269ed5291275a174a633b23dc92667de
Bug 833098: Part 1 - Remove dead code in xpfe directory viewer. r=Mossop

https://hg.mozilla.org/integration/mozilla-inbound/rev/c00426f6bc7d1a10e5609ac9530eb315d64d9d85
Bug 833098: Part 2 - Remove RDF service. r=hsivonen,Mossop,pike

Comment 32

[Dorel Luca [:dluca]]

Backed out 2 changesets (bug 833098) for depending on bug 1457749

Backout:
https://hg.mozilla.org/integration/mozilla-inbound/rev/d439053dcb4a8f88ca1be465cc01745b088852ed

Comment 33

[Kris Maglione [:kmag]]

https://hg.mozilla.org/integration/mozilla-inbound/rev/4cf33b7e542827c80c494eeea10c10c2721bf2d6
Bug 833098: Part 1 - Remove dead code in xpfe directory viewer. r=Mossop

https://hg.mozilla.org/integration/mozilla-inbound/rev/6f52281bccaa9f7624ec5bfac9e85e41f1928547
Bug 833098: Part 2 - Remove RDF service. r=hsivonen,Mossop,pike

Comment 34

[Cristina Coroiu [:ccoroiu]]

https://hg.mozilla.org/mozilla-central/rev/4cf33b7e5428
https://hg.mozilla.org/mozilla-central/rev/6f52281bccaa

Comment 35

[Mike Conley (:mconley) (:⚙️) (PTO: May 27-31)]

\o/