XSF Discussion - 2021-12-07

Good afternoon

😊

To me, most pages on the website seem slightly off-center (to the left). Is that just me?

Guus, now that you mention it.. the brand logo on my monitor does not align with the page (and that's centered)

Guus: tried adjusting your horizontal sync on your CRT?

har har

I'm not a web developer, but if I highlight some sections, then each section seems to be "OK". The right-hand sidebar is not filled fully, which leaves a lot of whitespace, which makes thing appear as if they're out of center, I think.

https://igniterealtime.org:443/httpfileupload/rOTFXL0c-bSzAQYa1wMN8lGkXLU/image.png

I'm unsure if this needs fixing, and if it does, how that is done. I thought I'd mention in the hopes that an actual capable webdeveloper (basically - anyone but me) has an idea. 🙂

Guus, I found some things which could be improved. I'll have a look this evening :)

Most excellent 👍

Observation: In e.g. pubsub disco#items listings, the jid of the pubsub service gets repeated a lot. E.g. as in https://xmpp.org/extensions/xep-0060.html#example-10 Feels ... redundant somehow. The 'jid' attr is MUST per '30 tho.

compress it!

Compression + encryption bad, remember? 😕

yes, I remember the discussion, I still think compressiing at least individual stanzas isn't harmful and would help in this case a lot ✏

yes, I remember the discussion, I still think compressing at least individual stanzas isn't harmful and would help in this case a lot ✏

and I think we either need to be sure *all* implementations switch to the safe way, or introduce a new compression mechanism that has very strong words about safety

also not all that excited about the additional memory usage compression needs

I am sure those can be allowed to turn compression off

It's not like you have to enforce compression on ppl :)

exi to the rescue

to continue repeating points from the last time we did this, fixed dictionary compression!

Daniel, got implementation?

no. i just wanted to make sure we brought up all the arguments from last time again

Zash, got money?

Daniel, 👍️

hmm how positive are we that we already brought up all the arguments? :)

Zash, I actually looked into fixed dictionary compression for a few hours and it has been unclear to me what libraries/tools actually support this

it looks like you can give zstd a starting dictionary. but emphasis on the starting

I thought zstd supported a fixed dictionary

i don’t know. my google search and reading docs didn’t bring anything up.

just have per JID dictionaries, and use some outer container format to distinguish by JID... maybe XML?

Or what about HTTP/3, it has streams!

Stream per JID pair?

Dreams of SCTP?

You could also do something like FunXMPP... except also writing it down in a XEP.

\1 → <message, \2 → <presence etc

Daniel already suggested EXI, so...

ejabberd has something like this for database compression iirc

Yeah, I heard of that.

concerning zstd: you can start with a fixed dictionary, and set a max size for the dictionary (and if that max size happens to be the size of your starting dictionary...)

that said, I don't feel compression per stanza offers full protection from the types of inspection attacks which are weaknesses for compression+encryption, though I have no proof either way

so, trade-off?

I have a half-finished compact xml representation scheme using a semi-fixed dictionary (only server can update) and forbids adding security related tokens to the dictionary; to be finished Someday™

trade-off what?

compression ratio/efficiency vs security protection

I think security is either yes or no

I think It Depends™

I'll choose 'yes', please.

the problem with naively compressing streams is that you don't have the option of not including security tokens in the compressed data - which is how most of the attacks work

but an immutable fixed dictionary should avoid that

> concerning zstd: you can start with a fixed dictionary, and set a max size for the dictionary (and if that max size happens to be the size of your starting dictionary. is there a chance that this will still mutate the dict though? theTedd

I honestly don't know, but I presume it wouldn't as that would almost certainly change the size

but I see you mean substituting entries, so I'm not sure, but I think not

right. yes I considered this 'hack' after reading the doc. but i wasn’t sure if anyone ever does this. or if it is somewhat supposed to work like this

> but I see you mean substituting entries, so I'm not sure, but I think not yes that's what i meant

Given that the docs talk about both a dictionary method and a prefix method, I would think the prefix method would be what you might call a mutable dictionary