On 2020-11-28, at 23:02, John Levine <johnl@xxxxxxxxx> wrote: > > It looks to me like the vast majority of sqlite applications use the C > library from sqlite.org and it's not clear that there's another full > implementation. This raises the question of whether the real > definition is the spec or the code. Indeed, it is hard to keep a spec document alive if there is only one implementation and that is moving fast. Compare the recent discussion of rsync, which (apart from a heroic reimplementation effort, which became necessary because of licensing issues, but seems to be stuck at protocol version 27, ca. 2004) is another one of these single-implementation standards. > There are a lot of widely used data format that the IETF and other > SDOs haven't blessed, like FITS which is a widely used format for > multidimensional numeric data, but whose definition is maintained by > the astronomers who originally designed it. The 800-pound gorilla in this space is, of course, HDF5. I don’t think they need the IETF (except that maybe somebody should tell them that text really is always UTF-8 these days). I think it would be an interesting research project to do a survey of these table-/array-oriented data interchange formats and derive a set of best practices and useful approaches, as well as possibly unexplored avenues. This would mesh well with various data modeling and semantic interoperability activities we have in IETF and IRTF. Certainly not something to standardize, and most certainly not https://xkcd.com/927/. Grüße, Carsten
