Re: Transforming hmtx table

Thanks Raph.

I generally agree with reducing non-trivial complexity.  But in the case of
lsb's that match the xMin, this is no more complex than reconstructing
loca.  The gains are indeed small, but now that all the big fish has been
fried, these look like low-hanging fruit.

b

On Fri, Apr 24, 2015 at 11:40 AM, Raph Levien <raph@google.com> wrote:

> David, thanks for dredging up that document, I couldn't find it myself
> because I misspelled it "htmx".
>
> I did consider it, but it seemed like it would add nontrivial complexity
> to the decoder, so didn't seem like it would be worth it. The document
> outlines some other potential transforms (cmap, etc), but basically in the
> tradeoffs between complexity (which includes decode time) and compression I
> ended up (after discussion with others in the group) only going after the
> biggest wins, which of course turned out to be glyf/loca.
>
> Raph
>
> On Fri, Apr 24, 2015 at 11:25 AM, David Kuettel <kuettel@google.com>
> wrote:
>
>> On Fri, Apr 24, 2015 at 8:56 AM, Levantovsky, Vladimir <
>> Vladimir.Levantovsky@monotype.com> wrote:
>>
>>>  Hi Behdad,
>>>
>>>
>>>
>>> The original MTX had pre-processing steps defined for hdmx and vdmx
>>> tables but I don’t think we did anything with hmtx. I remember Raph ran an
>>> estimate and concluded that hdmx/vdmx processing had ‘unfavorable’ gain
>>> /complexity ratio contributing little to overall compression so we decided
>>> to not have it as part of WOFF2 pre-processing step. I am not sure if we
>>> evaluated hmtx pre-processing at the same time, may be Raph can help to
>>> remember what, if anything, we considered w.r.t. hmtx table.
>>>
>>
>> Here for reference is a copy of the document that Raph originally wrote,
>> which captured the various optimization explorations:
>>
>>
>> http://wiki.font-compression-reference.googlecode.com/git/img/WOFFUltraCondensed.pdf
>>
>> There is a section on the related hdmx table, which Raph recommended
>> either just compressing or stripping altogether.
>>
>>>
>>>
>>> Cheers,
>>>
>>> Vlad
>>>
>>>
>>>
>>>
>>>
>>> *From:* Behdad Esfahbod [mailto:behdad@google.com]
>>> *Sent:* Friday, April 24, 2015 2:30 AM
>>> *To:* WOFF Working Group
>>> *Subject:* Transforming hmtx table
>>>
>>>
>>>
>>> Hi,
>>>
>>>
>>>
>>> Has the working group considered transforming the hmtx table to remove
>>> the lsb field completely, if the lsb is the same as the xMin from glyf
>>> table for all glyphs?  That will save two bytes per glyph for TrueType
>>> fonts before Brotli compression.
>>>
>>>
>>>
>>> behdad
>>>
>>
>>
>