Problems on Slack.com Topic is solved

[SLotman]

I usually access some groups on slack.com - but since 27.2 it keeps telling me that it is "offline, and it is reconnecting". Can't do anything on the site, because of that.

I don't know if it is something on their end, or if the 27.2 update broke something. What they do have is a online connection test, and that is failing in one of the steps:

https://slack.com/help/test

It fails for me on: "Web Socket (Flannel): Failed (Socket closed unexpectedly)"

If I access it on Chrome, it works. Previous Pale Moon versions before 27.2 were working too.

I also reported this to them - but since I have no clue what is going on (tried another useragent, tried disabling adblocks, private mode, etc - no effect), I'm coming here for help too.

Does anyone else in here uses slack? Do you have this same problem? Found a solution?

Anyway, thanks in advance

[billmcct]

Works for me.
Try cleaning cache and cookies.

Win 7 x64
PM 27.2.0 32bit

[Moonchild]

The test page includes a handy link to "tell them about it". You may want to use that, to see what the problem is.

Another person also mentioned problems with slack and it seems the connections simply get closed, indeed. It's a recent issue, apparently. Flannel [No compression] succeeds, so I'm gathering they are trying to use some form of compression which causes issues. Only Slack can provide additional info here.

[Moonchild]

Please check if setting network.websocket.extensions.permessage-deflate to false fixes your issue.

They may be using websocket compression in a way that falls outside of the extension spec (but which Chrome implements) and expecting Pale Moon to send extra parameters, specifically "server_max_window_bits" and as a result don't understand Pale Moon's request (lacking those parameters) while indicating the extension is supported (which it is and per spec), and their server script falls over as a result (best guess) not being tolerant to different implementations. The spec neither requires nor encourages this parameter to be sent.

See RFC7692, authored by Google Inc (no surprise).

A client MAY include the "server_max_window_bits" extension parameter
in an extension negotiation offer. This parameter has a decimal
integer value without leading zeroes between 8 to 15, inclusive,
indicating the base-2 logarithm of the LZ77 sliding window size, and
MUST conform to the ABNF below.

[billmcct]

network.websocket.extensions.permessage-deflate is set to false in my settings. Don't remember ever changing it though.

[Moonchild]

network.websocket.extensions.permessage-deflate is set to false in my settings. Don't remember ever changing it though.

That explains your result, then
It may have been changed by a tool/extension if you didn't do it yourself by hand. I'd gather some people would make this part of some "optimization" efforts.

[billmcct]

Seems to be the case. Shows User Set.

[Moonchild]

Contacting slack support I got the same form reply "we don't support Pale Moon bla bla" so I relayed the details to them. I hope they will listen and actually support a standards-compliant web as opposed to a monoculture web.

Disabling the extension to websockets cleared up the problem and slack is fully functional, so I'm pretty sure it's just their assumption that all browsers use the out-of-spec Chrome "extras" and falling over if not.

[SLotman]

Yeah, same answer here, "We don't support Pale Moon bla bla bla" :/

But setting network.websocket.extensions.permessage-deflate to false really solved it! Thank you!!!

[app103]

Seems to be the case. Shows User Set.

In my case, it wasn't user set. The default setting was true.

[Moonchild]

I've had some more discussion with the slack people and it seems the RFC actually has a conflict in its wording. Slack sends a parameter in their response that is allowed but not required for a client to implement. Pale Moon doesn't have this implemented and is asked to use an unknown (unsupported) configuration. According to the RFC, the client should close the socket connection in that case. So we're both adhering to our side of the spec

I'll be implementing these optional parameters to work around this spec issue in the next release of Pale Moon, and have asked the spec author (some person at Google ) to have another look at it (apparently this simple fact slipped by everyone through over 20 drafts and made it onto the standards track...). I've asked slack to look at an interim solution to allow Pale Moon users to connect, despite this spec issue (and actually, compression over SSL this way (re-init compression for every message, see CRIME) may be a security issue as well if they haven't taken this into account in their server implementation).

[SLotman]

Greeeeat. Now Slack just blocks Pale Moon entirely :/

Edit: just submitted a complaint over there about it.
Edit 2: I can now bypass that screen by using the following user agent-string

Code: Select all

Mozilla/5.0 (Windows NT 6.1) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/55.0.2883 Safari/537.36

[RJARRRPCGP]

Looks like a webmaster retaliated by persecuting Pale Moon!

[New Tobin Paradigm]

Oh wow... Like.. wow.

[h0w4rd]

Bloody hell... I was using it yesterday without any problem

[Moonchild]

I sent them another reply on their ticket discussing this. What the hell.

[h0w4rd]

Please how can I set the user agent only for specific page? And is it helpful?

[Moonchild]

And on the other side of the fence, reporting the obvious spec conflict to its author, I was met with a lot of resistance.

EDIT: Looks like they finally understood the issue.

[SLotman]

Look at the "marvelous" response I got today from my complaint:

Hi there,

Thanks for taking the time to write in today. Unfortunately we no longer support Pale Moon, however, this might be reconsidered in the future if this browser meets specifications in-line with our security practices.

The four supported browsers are the latest versions of Chrome , Firefox, Safari, IE / Edge:

Chrome (and Chromium) version 51 and newer
Firefox version 47 and newer, plus version 45
Safari 9 and 10
Microsoft Internet Explorer 11
Microsoft Edge version 14 and newer

I'm very sorry to be the bearer of this news to you, however the main reason for this is since we take security and product quality very seriously. We want all users to benefit from the latest features, security, and bug fixes.

My suggestion would be to use one of the supported browsers listed above, I hope that helps and sorry again.

Why do I even bother to post something over there...?

Please how can I set the user agent only for specific page? And is it helpful?

In this case, just type "about:config" on your URL bar. Right click on the list that will open below, and add a new string value. That string would be: "general.useragent.override.slack.com" and the value, the one I posted above. You can do that to any website, just replacing "slack.com" with it's URL.

The benefits of that is obvious: if a site is doing "browser sniffing", aka reading the name of the browser to serve a specific feature (or in case of Slack, to block the entire site) - you can bypass it to get better results.

[Moonchild]

Serious accusations there. I can't for the life of me understand why they would claim "we're not upholding specifications in line with their security practices" when, in fact, we have tighter security than most browsers, and on top, I pointed a potential problem out to them with repeat compression over SSL (e.g.: CRIME) involved, to which they didn't respond.

This is ridiculous.