Iljitsch van Beijnum
iljitsch at muada.com
Sun Apr 8 11:16:22 PDT 2012
I just had a look at http://tools.ietf.org/id/draft-rfc-image-files-03.txt
I think this is a spectacularly bad idea. This way, we would have all the downsides of PDF, all the downsides of formatted ASCII text, and lose the advantage that we only have a single file.
This also locks us into pagination even deeper, while IMO that's something we should move away from, because the number of times when US letter pagination is convenient are dwarfed by the number of times when it is not.
There are lots of tools that read and write PDFs, but that doesn't mean it's necessarily reasonably possible to round trip an image from an image editing program to PDF and back without losing a lot of information. For instance, the image may be rasterized so what was a vector image is now a high resolution bitmap. Or logical structures are broken up in lots of small structures but the relationship between them is lost.
I don't think these issues are easily avoidable without blessing a specific tool. And tools tend to go away on decades timescales. Therefore I think that if we want to allow images as authoritative parts of RFCs (which I still have to see a good case for) this needs to be relatively low resolution bitmaps in an open format, like PNG. Bitmaps have the advantage that they are very simple structures that are easily manipulated with simple software, not unlike ASCII files.
Finally, let me observe that a lot of people here are applying a tinkering/hacking mindset, trying to get some benefits with modest changes. I don't think that's the right approach here. It's perfectly fine to make very big changes if that gives us what we need. Small steps are a bad thing here, that means that 20 years from now the tools that scrape the RFC repository need to support many variations that were around for only a short time rather than just "old" and "new".
More information about the rfc-interest