NDT-DEV email list created

[Athanasios Douitsis <>]

Hello Aaron, thanks for the reply!,

Re: websockets: I think that I might be eventually able to come up with a small patch, but it may be slightly more complex that the one liner I envisaged. No matter though, yes, I'll issue a pull request if and when I do. My tests already play and detect failures seemingly correctly, but I'll try to think about possible corner cases before rushing it out the door.

Re: firefox+websockets+workers: Sweet indeed, provided that it's accurate, which I don't know yet. But it works and the measurements . I am not using the NDTWrapper in ndt-wrapper.js but I rather have created a promise which invokes the worker in ndt-wrapper-ww.js, regardless of browser type. Works and produces similar results in firefox and chrome, haven't tested elsewhere. Do you have a formal/semiformal way of ascertaining that a measurement is accurate enough?

Kind regards,

Athanasios

On Thu, Oct 22, 2015 at 7:57 PM, Aaron Brown <> wrote:

Hi,

On Oct 21, 2015, at 11:06 AM, Athanasios Douitsis <> wrote:

Hello,

In case one desires to submit a small patch (e.g. for the ndt-browser-client.js) under HTML5-frontend, would a pull request on https://github.com/ndt-project/ndt be appropriate?

You stated it wouldn't be easy but if you do come up with one, that's the easiest way.

I have run across an mlab ndt server that for some reason does not respond normally to the websocket version of the protocol. While this sounds bad, it actually gave me the opportunity to realize that, upon creation of the websocket in ndt-browser-client.js, the onclose and onerror events are not handled, so if the connection is dropped abnormally or any other sort of error occurs, the application using NDTjs doesn't learn it and can hang in there indefinitely.

If I may be permitted, I'd like to submit a small patch to fix that.

Also, another question, in my tests I am using Workers under Firefox quite successfully. It seems that the bug saying that WebSockets were not available in Workers (*) is nowadays fixed. Anyone aware whether there is any issue concerning the accuracy of measurements? Because as far as I can tell it works beautifully.

Sweet. I guess it's a more recent version. If you'd be interested in patching that issue, we'd be happy to accept it. Otherwise, I'm happy to take a look at fixing it.

Cheers,

Aaron

(*) https://bugzilla.mozilla.org/show_bug.cgi?id=504553

Thanks,

--

Athanasios Douitsis

--

Athanasios Douitsis