[nanomsg] Re: Review of mangos

  • From: Paul Colomiets <paul@xxxxxxxxxxxxxx>
  • To: "nanomsg@xxxxxxxxxxxxx" <nanomsg@xxxxxxxxxxxxx>
  • Date: Wed, 7 May 2014 12:15:45 +0300


On Wed, May 7, 2014 at 8:29 AM, Martin Sustrik <sustrik@xxxxxxxxxx> wrote:
> Hash: SHA1
> On 06/05/14 18:52, Gonzalo Diethelm wrote:
>> Line 554: why is it subscribing to every possible message by
>> default? This could be a bad idea...
> Yes, we've seen this in real world. The subscriber gets hit by a ton
> of traffic before it's able to unsubscribe. It may never recover from
> the blow.

I think it's because nanocat does it. And the reason is because
nanocat is just a simple utility to test how things work. And the
simpler it is to use, the better. I.e. what do you think default
should be? No subscriptions? The ``nanocat --sub`` is useless without
subscriptions (note, unlike with raw sockets, you can't add
subscriptions on the fly in nanocat). Require ``--subscribe``
argument? I think it's too verbose for simple cases of debugging.  All
in all, if you really connect with nanocat to some huge production
system, with millions of record per second, you are probably smart
enough to add ``--subscribe`` argument.


Other related posts: