Binary diffs and largefiles

dukeofgaming dukeofgaming at
Fri Sep 21 10:26:42 CDT 2012

The reason I was looking into this is because we need to implement a better
mechanism to distribute release updates (this is for an embedded system
that runs on a single core OMAP platform [say, something of the
capabilities of a good single-core smartphone]). These updates would most
probably be several ISOs.

So, just asking from the inverse perspective: do you think Mercurial would
be suited as a bsdiff alternative to some degree (because of the memory
consumption)?, and even maybe for distributing updates?



On Thu, Sep 20, 2012 at 1:12 PM, Matt Mackall <mpm at> wrote:

> On Thu, 2012-09-20 at 09:40 -0500, dukeofgaming wrote:
> > Hi,
> >
> > I've been doing a little research on binary diffing and patching and I
> was
> > wondering if this was something that made sense for largefiles and how
> > binaries are stored.
> >
> >
> >
> a) Non-largefiles Mercurial already does "binary diffs" internally that
> are probably comparable to bsdiff just fine for files up to hundreds of
> MBs
> b) One big reason -not- to use plain old Mercurial for large files is
> because it can't handle creating/applying deltas on files that are near
> the size of memory, but bsdiff uses _much more memory_
> c) The most interesting largefiles (JARs/zips/images/movies/audio) are
> not compressable/diffable anyway
> --
> Mathematics is the supreme nostalgia of our time.
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <>

More information about the Mercurial-devel mailing list