XSF Discussion - 2018-09-18

hi sslh was configured but it log in syslog, it was followed this manual: http://william.shallum.net/random-notes/sslh-configuring-logging-logrotate-and-logwatch and it still log in syslod, sslh was restart and same

I mean the answer is you configure them however you want

You need to understand ips and ports and what sslh does though, you aren't going to find a copy paste Tut for it

peter: nice so 'I'm going to generalize about all federated systems then come up with one that isn't federated because I haven't considered how to solve those problems in a federated system'

moparisthebest: your tutorial was good... but it suck... simply step by step... and you have to remove , from line 16... Imagine, today you was online... person want install ssllh, he see your tutorial, start scratch in hear... make everything correctly... after some hours to see how to make it work " without step by step he need to lost hime time " and when he run sshl -F /etc/sslh.cfg it will comment " error in line 16 " he will start again scratch head hair and agai other 1 day to configure how to repair the problem. so 2 days to install, configure and run sslhd. but here come true life. what happend if he dont find you moparisthebest... so he has to wait more days... maybe 2 days more until you return... or maybe 1 month... or maybe 1 year... becouse when person dont know how to run somethink. he has to wait untill someone help him... so imagine he has to wait 1 year untill someone or you appear help him. this is 1 year to remove one " simply SHIT , " if he has goodluck he cab wait 1 year, and if he dont habe goodluck it can be 2 years... and more... so for some manual that can work 10 min. he has to wait 1 year or more years... to someone help him. moparisthebest today you can have internet, tomorrow do you know if you are gonna have internet... you can " i pay money " - > piu piu... you can money but when government say no more internet there will be no more internet...

moparisthebest: and who is the not federated server...

moparisthebest: is this some kind of indirect you want my help for federated system. if i can i will help.

moparisthebest: sslh is program for linux " unknown if it work in macos " that help peoples with limited ports. all other ports blocked only open 443 and 80 -> only for viewing web sites... sslh enter here and safe peoples... limited port and firewall... have noce day.

What in the world kind of drunken rant was that

What in the world kind of drunken rant was that

I had some though on OMEMO and related metadata, the problem with OMEMO (and other e2ee solutions) is that they are only encrypting the body element. I was wondering if we couldn't also encrypt, using the same key, the other value of the other attributes of the message (by adding a namespace to those elements for example). For SIMS we could then encrypt <media-type>, <name>, <thumbnail cid…>.

This will still leak some "structural metadata" but at least the content should be protected. And it should be fairly easy to implement it in clients (we have to check for the retro compability).

its a bit more work then just encrypting everything

there are elements you dont want to encrypt

yes :)

and these have to be specified :)

as daniel often said, there is probably no one against full stanza encryption, just no one did write it up yet

that's why I'm saying that we should put a specific tag or namespace for those specific tags

the problem with full stanza encryption is that it requires to hack the parser behavior and thing that should be encrypted and things that shoudn't can't be on the sams XML depth anymore

full stanza encryption is specified in 0373 and i think 0200

also a question was, it should be specified what happens if i find a tag inside the encrypted payload, and outside

Or what happens if there is a tag outside that influences the inside. For example a message correction tag

i think we should probably ignore all outside tags except a certain whitelist

but then the problem is, if new xeps arise we have to update that list

I wonder which tags are relevant to clients && must not be encrypted

eme for one

its not about client, its about server

hints for example

sure

but the server shouldn’t care about the encrypted payload

and the client shouldn’t care about the unencrypted payload (in general)

you know where I’m getting at?

> i think we should probably ignore all outside tags except a certain whitelist Thats what I proposed years ago

i.e. the whitelist will probably be encryption-metadata like EME, and everything else can be ignored by the client.

I can only think of stanza and origin id

or stanza-metadata, hm, yes.

Eme is only relevant _before_ the decryption

yes stanza-id

And can be thrown away later

but this is added by the server anyway

but should not be ignored yes

https://news.ycombinator.com/item?id=18013275

?

MattJ thanks for the comment :)

https://beesbuzz.biz/blog/2535-ActivityPub-hot-take

Instead of retrieving all history from MAM, I'd like to ask the archive the last message for each party I conversed with to build a chronological index, and then progressively retrieve history when going into one of them.

Has anyone had a use case like this? If so, how did you solve it?

ralphm, I was planning to write an extension for that. MAP (Message Archive Preview) :)

Sounds great

It seems clients are favouring the per-contact sync, instead of sync-everything

I could imagine maybe a boolean field in the MAM Data Form that indicates you only want one for each party?

Potentially, yes

MattJ: indeed. Also in our case, we don't really have a roster

You have people on your (phone) contact list, and then other sources of things to talk to, like groups and non-people.

By the way, I love the MAP backronym.

:)

Fancy working on the XEP?

I do believe Kev has talked about some kind of summary like this as well

Reality is, I probably won't get to it for a couple of weeks at least

It's much better than the CSI backronym.

We probably are fast-moving on this, but that doesn't mean I can't work on a standard for it and move to that later

ralphm, do you request the last message for every contact in the roster on start?

my question is about how do you solve the problem that you cannot know what the last conversation was in a multi device env

so how do you knwo which jids to sync

lovetox, all

Last message per contact

since a certain mam-id

obviously, otherwise this would be very expensive

That sounds massively more expensive

Keeping the last ID per roster entry in a cache seems doable

note that ralphm was specifically talking about not really having a roster

No, no roster

You know I'll cry if you design a thing that requires a SQL RDBM

`SELECT MAX(uuid),* FROM contacts, messages WHERE contacts.jid = messages.jid;`

> MAX(uuid)

what does that mean no roster, you talk only to a single jid? because if you have knowledge about more than one jid in your application you have a roster, i dont see how it is relevant if this roster is kept on the server or on the client

that’s not how it works.

My thinking was that you do a MAM request to get all message history, but instruct it to only return one entry per unique other entity you've contacted (other user, room, whatever)

jonas’: stop spoiling my fun

lovetox, it is relevant for the server-side MAM implementation whether it knows which JIDs are relevant or whether it has to assume they all are

No nice things allowed.

ralphm: that might explode if you have chatted to many entities. Or if you were bot-flooded.

lovetox: in our case, instead of a roster, we have a native mobile phone address book, and retrieve matches from the server. Not unlike apps like WhatsApp do.

Ge0rG: well, you can paginate this too, no?

Ge0rG: or are you worried about the server side complexity of 'the last one' per contact?

ralphm do you store the messages on the phone?

lovetox: cache, yes

ralphm: both, actually

but you definitely don't want to retrieve all messages at once either, most of the time you only need history for a contact when you start chatting

ralphm, I’m not sure how many messages you are expecting between reconnects and how many contacts you expect people to have, but if the message load is "regular", you might get less traffic with normal "MAM since last connct" than what you propose

jonas’: not on a reinstall

because the one is O(number of contacts ever ever spoken to [monotonically increasing]) and the other is O(number of contacts actively sending messages times average message rate per contact times offline time [approximately constant])

or when you didn't charge your phone for a month.

ralphm, okay, that makes sense

but on a reinstall, I wouldn’t worry too much and just do a MAM query for each phonebook entry

eh no

do you expect huge phone books?

jonas’: there is an upper bound due to the typical history age timeout, so you'll only get the contacts you chatted to in the last 14d or so

if you have implemented something like that, im very interested how you did it, i try to find a way to backfill the history since a month, and all i can come up is that its not possible without lossing the order of the messages

Ge0rG, depends on the history model on the server side, really

Ge0rG: indeed. I could imagine retrieving the last 7 days of history, but you still want an entry for people you talked to longer ago

And then if you go into one of those chats, you can still backfill that contact

ralphm: so you are also asking for indefinite storage of all JIDs you had a chat with?

The thing is that you don't want a full history sync when you reinstall an app or switched phones or whatever

Ge0rG: might still be definite, but still different from a contact list. Contact lists change and also you want the chronological order

The server-side complexity of getting a list like this is an implementation detail. It could be part of the storage model, or you can build and cache an index based on the actual full history.