- From: Dirk Pranke <dpranke@google.com>
- Date: Wed, 5 Aug 2009 12:20:20 -0700
- To: Jonathan Kew <jonathan@jfkew.plus.com>
- Cc: Sylvain Galineau <sylvaing@microsoft.com>, John Daggett <jdaggett@mozilla.com>, www-font <www-font@w3.org>
On Wed, Aug 5, 2009 at 1:29 AM, Jonathan Kew<jonathan@jfkew.plus.com> wrote: > On 5 Aug 2009, at 02:18, Sylvain Galineau wrote: > >>> Others seem to view EOT-Lite as a stepping stone format that would be >>> followed by a better .webfont/ZOT/something-else format. But another >>> new format would need to offer a big marginal advantage to offset the >>> disruption supporting yet another format would cause. >> >> I lean towards being one of them. What kind of specific disruption(s) are >> we >> Talking about though ? > > One concern I have is that if we all implement EOT-Lite (by whatever name) > now, it will be more difficult for a new and better format to gain > acceptance, and so we may end up with an inferior format in the long term > for the sake of short-term gain. If we introduce a format such as > .webfont/.zot/etc at this point, there's a strong incentive for everyone to > get on board; this will be reduced if there's EOTL as an available, > interoperable solution even if it is technically less optimal. This argument is theoretically sound, but I would prefer to see something more concrete. In particular, (a) what features do we think are missing from EOTL that we'd like, (b) can we implement them compatibly inside EOTL somehow? (c) is there something other than (a) and (b) that causes us to not like EOTL? The only real features I've heard suggested are compression and subsetting, and better metadata support. I haven't thought about this at all, but given that EOT had MTX support, I would assume we could also add it to EOTL compatibly, no? If not, you can certainly compress the whole file using whatever algorithm you like, just like we do with gzip/deflat today. Subsetting can happen at the OTF level, right, so that can also be done? That leaves metadata support. I haven't heard anything that says that the desired metadata fields can't be put directly into the OTF/TTF files. Am I mistaken? It seems like since OTF already has extensibility mechanisms, we should just be using those, rather than trying to recreate them in a different layer/wrapper. But perhaps implementors would object because they'd have to parse the OTF format rather than just handing it to the O/S to render. That would seem to leave us with some objection to the wire format (not XML, not MIME, not .ZIP, whatever). Frankly, I don't care much about that at all when traded off against simplicity of implementation and compatibility gains. It's not like OTF/TTF are human-parsable either, and it's not like SVG has gained widespread adoption because it is ... Are there other concerns? Again, I am not necessarily voting in favor of EOTL, just trying to understand what the "newer, better" format would really offer that couldn't be achieved. -- Dirk
Received on Wednesday, 5 August 2009 19:21:03 UTC