----- Original Message -----
> Hi -
>
> These last two pcp-archive-related bugs made me search around for
> formal documentation of the archive file format, but I couldn't find
> any. If we are to fix some of these problems, we'll need to come up
> with some specs, so we have more than de-facto ("whatever the current
> libpcp, libpcpi, pmlogger, pmlog* functions each accept/create") since
> those will change.
Some documentation about the metadata and temporal index file formats
sounds like a wonderful idea. The data format is described via proxy
in pmFetch(3) and in the books, where pmResult formats are detailed,
but these other aspects are not documented AFAIK. And there are some
aspects of the data file (label, mark records) which are not deeply
discussed there too.
> (That is, unless we imagine doing something drastic like switching
> away from these files as the preferred form.)
We might imagine it, but always we will need backward compatibility;
so whatever happens this documentation will be extremely useful. If
diagrams would help (svg or what-not) consider the Programmers Guide
as a possible home for the docs too. Very happy to offer my limited
svg drawing skills (used for other PG docs) if sketch diagrams could
be made.
thanks!
--
Nathan
|