So actually providing details to this bug report rather that having everyone read through a mailing list thread...

The speed up is done by only extracting the new folders in the db archive on -Sy operations as regular updater will only have a few package differences.

So this seems a waste of time as we are going to implement a full tar backend one day... ( FS#8586 ). Then no extraction would need done at. Using tar based local db would also speed up -Q operations.

rebase is a perl script written by Xyne which does the same :
http://bbs.archlinux.org/viewtopic.php?id=65313

Xavier: Personally I am not a big fan of millions of 3rd-party scripts. (I haven't even heard of rebase before, it is not in pacman-contrib. But it is good to know that such script exists, so thx for mentioning it.) And the existence of rebase script should not affect our decision on pacman (implement this or not), for example, yaourt has plenty of features, but we don't suggest pacman users to use that instead of "core" pacman...

I agree with Allan here, we should implement  FS#8586  instead of this request [or use ~sqlite :-P].

I am not a big fan either.
But if we reject this bug because of future plan for pacman, it is useful to know this script exists in the meantime :)

I've discussed this FR with the reporter on irc, and my opinion has been slightly changed:
If  FS#8586  is not implemented soon (in 3.3, for example), we may want to implement this FR first, because we can reach _notable_ speed-up at low "coding cost".

However, there is a draw-back I must mention:
If the sync database becomes corrupt, but the package directory remains valid, -Sy may not fix this sync database corruption, which is weird. (Frugalware guys kept the old behavior with -Syy to workaround this.)

"If possible please implement it in the 'archlinux' pacman version."

Not possible. Unless if you provide a patch, then it will be considered.
Obviously one which applies to archlinux pacman, master git branch.

I see the preliminary commits for this made it in master:
http://projects.archlinux.org/pacman.git/commit/?id=14ab02e2
http://projects.archlinux.org/pacman.git/commit/?id=3dc87851

But not the final one:
http://code.toofishes.net/cgit/xavier/pacman.git/commit/?h=working&id=e7aa4cc6

Indeed, I sent that to the ML but Dan never ACKed it.
I got a very useful review from Henning Garus though and re-submitted a fixed patch.

Hmm... Dan broke the patch: http://projects.archlinux.org/pacman.git/commit/?id=d2dbb04a :P I had a quick look but I do not understand some of the changes Dan made without looking deeper so I could not rebase the patch.