pcp
[Top] [All Lists]

Re: __pmDiscoverServices()/pmfind Enhancements

To: Dave Brolley <brolley@xxxxxxxxxx>
Subject: Re: __pmDiscoverServices()/pmfind Enhancements
From: fche@xxxxxxxxxx (Frank Ch. Eigler)
Date: Wed, 14 May 2014 14:39:47 -0400
Cc: pcp@xxxxxxxxxxx
Delivered-to: pcp@xxxxxxxxxxx
In-reply-to: <5373ABAD.4030300@xxxxxxxxxx> (Dave Brolley's message of "Wed, 14 May 2014 13:45:17 -0400")
References: <5373AAA7.70004@xxxxxxxxxx> <5373ABAD.4030300@xxxxxxxxxx>
User-agent: Gnus/5.1008 (Gnus v5.10.8) Emacs/21.4 (gnu/linux)

brolley wrote:

> [...]
>> Thoughts? More ideas while we're at it?

(All of the above make sense.)

> Heh -- just thought of another one.
>
> In the current API, we know which service we're trying to discover
> (e.g. pmcd). Perhaps once it is known that something is listening, we
> should also try to connect as an actual client of that service
> (i.e. go through the handshake). [...]

The complete handshake can need more information than just the
hostspec (ip address:port) string, like ssl and authentication info.
We'd have to find some way of passing that info into the discovery
function.  Plus, we know that libpcp is not as thread-safe as formerly
believed, and making concurrent trial connections might rely upon the
problematic code.  I suggest leaving this aspect in the hands of the
pmDiscoverServices client.

By the way, the pmDiscoverServices.3 man page should include some
examples of output, to make concrete the generic "URL" term.


- FChE

<Prev in Thread] Current Thread [Next in Thread>