- From: Jonathan Kew <jonathan@jfkew.plus.com>
- Date: Fri, 7 Aug 2009 20:57:20 +0100
- To: Oliver Rigby <oliverrigby@gmail.com>
- Cc: Tal Leming <tal@typesupply.com>, www-font <www-font@w3.org>
On 7 Aug 2009, at 20:15, Oliver Rigby wrote: > On Fri, Aug 7, 2009 at 6:47 PM, Tal Leming<tal@typesupply.com> wrote: >> The structure of the XML is more extensible than a binary table. > > It is. > > I just don't think licensing data is sufficiently complex that it > needs such an elaborate extensible structured data format - I simply > don't think it justifies the extra complexity over a pure binary file > format. If we do a pure binary format, we'll be debating exactly what to include or exclude for the next N months, where N is a painfully large number. Using the more flexible XML approach allows us to avoid much of that pain. > > And if it _is_ necessary, I don't see why we shouldn't go the whole > hog and make the whole format, including the header, XML-based, > therefore gain the advantages of XML for the whole format - a format > that would be better structured, human readable, and be more > extensible and future-proof. The header and table directory are binary for the sake of efficiency: accessing the actual font data is on the critical path for browsers when they are rendering pages, and people complain about milliseconds. The binary header can be handled faster, and with far less code, than an XML equivalent. Accessing the metadata is not time-critical, so here we can favor flexibility over raw machine efficiency. > > Personally, I think this is a middle-ground compromise which gets the > worst of both worlds. As suggested above, I think many of us consider that it offers the *best* of both worlds. > By making it binary, it's not human-readable and > you need a specialised tool to make the fonts. By including XML, font > makers need XML parser and read an XML doctype, The alternatives are either a rigidly-defined binary format that doesn't need "parsing", or some other flexible format that still needs parsing, and if it's an ad hoc "simple" syntax then it's more likely there will be anomalies in parsing than if we use XML, for which there are mature, well-tested parsers readily available. > as well as increasing > the barriers to reading/writing metadata which may discourage people > supporting it at all, which would be the opposite of the intention. No matter what format we come up with, there needs to be a tool (or several tools, for different types of workflow) to make it easy for people to create. My sfnt2webotf command-line tool allows you to specify an arbitrary XML file to embed as metadata, but I expect someone will build a user-friendly WebOTF Metadata Editor for those who prefer that kind of interface. JK
Received on Friday, 7 August 2009 19:58:08 UTC