From practice - performance of clients and of servers too.
From emotion - it uses Web technologies.
From some logic maybe - if they are doing something new, then why not distributed architecture like Tox (at least identities not tied to servers), and if they choose something architecturally similar to XMPP, why not use XMPP.
However, emotion again, I really like Matrix APIs, these are definitely designed to be used by anyone at all.
It means it’s a robust well-tested protocol (referring to HTTP)
XMPP by now is no less well-tested.
Average company firewall: Allow 80 Allow 443 Allow 53 to <internal DNS server> Deny to any
Average company firewall shouldn’t allow 80 and 443 to outside anyway.
Anyway, that could have been a fallback, it’s the only way instead.
Doing an IM over TCP I can understand. VoIP signalling over TCP is not serious.
What’s the better solution?
Look at Retroshare. In this particular regard (not its whole model of security, which is seemingly not good, but I’m not a specialist) it does things right, I think.
Yeah it has a lot of problems, but all the things you listed are the least of it.
And which are not in your opinion?
Still better than anything else.
Still not better than XMPP, so factually wrong. =)
Less resource-heavy than Matrix, doesn’t have the “store everything from your every chat” feature and thus requiring less space, more mature, very easy to set up.
I got what you meant. Anyway, if it’s a company network, then they can, you know, allow something else.
Peer to peer is also a non starter.
That was in response to you asking how to do things without PKI, so I referred you to Retroshare as an example of using something like web of trust to that end.
P2P is irrelevant here. What does email have to do with this? Do you mean federation as in having servers, as opposed to distributed model? Do you mean identities being tied to servers?
And also why would that be “a non-starter”? Old Skype was P2P, using central servers for authentication only. I think we all agree it worked very well.
If you mean that it’s hard - I agree, I love to blabber about P2P solutions, but these are harder.
(Say, since old Skype people got used to downloading their history on a new device, which didn’t always work, but that can be solved by supernodes\servers to store and forward encrypted data with that history, a bit like Freenet. Only the person who can design something like that is definitely not me.)
What’s so good with XMPP?
What the other user said, plus having lots of good clients.
In general with XMPP thanks to the extension model (administrative one) good and bad things have already been tried, some discarded, and there’s a specific set of XEPs making it a very usable protocol supported by all relevant clients.
What’s wrong with Matrix?
From practice - performance of clients and of servers too.
From emotion - it uses Web technologies.
From some logic maybe - if they are doing something new, then why not distributed architecture like Tox (at least identities not tied to servers), and if they choose something architecturally similar to XMPP, why not use XMPP.
However, emotion again, I really like Matrix APIs, these are definitely designed to be used by anyone at all.
Oh no! Web based protocol! Not stability, ease of debugging, less block rate, and easy SSL protection! The horror!!
What does this even mean in the context of data you’d transfer in Matrix?
Ease in which context? What’s so much harder to which you are comparing it?
Are you certain that something TCP-based gives that? Latency sucks too.
PKI is crap. Just saying. Easy and wrong.
Nobody said that.
And such an esteemed thing as Gnutella uses Web technologies.
I just don’t like it. It’s my opinion. Just as you have yours.
It means it’s a robust well-tested protocol (referring to HTTP)
It’s a robust, well tested, and well known protocol.
Average company firewall: Allow 80 Allow 443 Allow 53 to <internal DNS server> Deny to any
What’s the better solution?
Yeah it has a lot of problems, but all the things you listed are the least of it. Still better than anything else.
XMPP by now is no less well-tested.
Average company firewall shouldn’t allow 80 and 443 to outside anyway.
Anyway, that could have been a fallback, it’s the only way instead.
Doing an IM over TCP I can understand. VoIP signalling over TCP is not serious.
Look at Retroshare. In this particular regard (not its whole model of security, which is seemingly not good, but I’m not a specialist) it does things right, I think.
And which are not in your opinion?
Still not better than XMPP, so factually wrong. =)
By firewall I mean outgoing. And XMPP is kind of a non-starter.
Peer to peer is also a non starter. You have to have some kind of email-like structure.
What’s so good with XMPP?
Less resource-heavy than Matrix, doesn’t have the “store everything from your every chat” feature and thus requiring less space, more mature, very easy to set up.
Hm. How’s E2EE?
I got what you meant. Anyway, if it’s a company network, then they can, you know, allow something else.
That was in response to you asking how to do things without PKI, so I referred you to Retroshare as an example of using something like web of trust to that end.
P2P is irrelevant here. What does email have to do with this? Do you mean federation as in having servers, as opposed to distributed model? Do you mean identities being tied to servers?
And also why would that be “a non-starter”? Old Skype was P2P, using central servers for authentication only. I think we all agree it worked very well.
If you mean that it’s hard - I agree, I love to blabber about P2P solutions, but these are harder.
(Say, since old Skype people got used to downloading their history on a new device, which didn’t always work, but that can be solved by supernodes\servers to store and forward encrypted data with that history, a bit like Freenet. Only the person who can design something like that is definitely not me.)
What the other user said, plus having lots of good clients.
In general with XMPP thanks to the extension model (administrative one) good and bad things have already been tried, some discarded, and there’s a specific set of XEPs making it a very usable protocol supported by all relevant clients.