I use NN_PRINT_ERRORS and the statistics. But I don't use any crufty ESTP
stuff.
Matthew.
On Apr 12, 2016, at 7:03 PM, Garrett D'Amore <garrett@xxxxxxxxxx> wrote:
I’ve been looking at libnanomsg, and the statistics socket publishing
facility is just one more bit of complexity that I think we might be better
off without. The documentation for it declares:
The output of both debugging facilities (NN_PRINT_ERRORS, NN_PRINT_STATISTICS)
is intended for reading by a human and a subject for change at any time (even
after 1.0 release).
The NN_STATISTICS_SOCKET facility is intended to be used by system
administrators to debug network and application issues. The ESTP format is
stable. But the nanomsg support of it is experimental and is subject to change
or removal until 1.0 release.
And from what I can tell, ESTP is not widely adopted. There is a collectd
plugin for it that I found but I’m not sure if that is used.
I’d like to remove this; api access to statistics should be available via a
regular API, which applications could then publish (if they choose to) via
whatever mechanism they want. I don’t think we need libnanomsg to know the
details of ESTP in particular.) Also, the application could then chose both
where and how often to publish such data, which seems like a far better
solution.
(The reason I am looking at this is in response to some gmtime_r related
machinery and hackery made for Windows & MinGW. I was fairly shocked to find
that we actually are formatting timestamp data!)
- Garrett
