#ls2 (irc2p) | #ls2 I2P Protocol Development Discussion - Next Meeting Monday, October 9th 6:30 UTC

eyedeekay Hi everyone welcome to the meeting

eyedeekay I'm having connectivity issues so let's try and keep things on topic

eyedeekay Who has topics for the meeting?

eyedeekay I'll start by adding "NetDB Change Summmary"

eyedeekay Who else is here?

not_bob Domains, can they auto-expire after disuse? Say a sunset of six months or a year?

not_bob Or, should domains be a forever thing?

eyedeekay Thanks not_bob sounds good. Anybody else got one a topic?

orignal hi

obscuratus hi

orignal I'm busy today

eyedeekay Hi orignal, hi obscuratus

eyedeekay OK that's fine orignal

obscuratus Update on RI secure handshake?

eyedeekay IIRC we tabled that one for a later date because of the SSU2 change it would require

obscuratus OK, if orignal isn't available, no need anyways.

orignal next week please

eyedeekay I'll be on a plane next Monday, can we do Tuesday instead?

orignal I also want to bring an issue with false positive peer tests

orignal fine

eyedeekay OK

eyedeekay On to the topics, let's start with you not_bob

eyedeekay You explained the general idea but would you like to give us a little more about the motivations, benefits, and drawbacks?

not_bob Many domains get registred, used for a very short time and then not used again.

eyedeekay This is for #1 domain name expiration

not_bob I've seen it happen countless times. Where the site is only for days, or a few weeks and is never to be seen again. But, then it eats the domain name forever.

not_bob That's my main modivation.

not_bob But, drawbacks include trust.

eyedeekay Squatting by any other name, really

not_bob It's good to know that xxx.i2p is the same xxx.i2p you visited three years ago.

not_bob With what I propose, so long as you stay active with your domain, then you keep it forever.

not_bob With a six month, or one year grace period.

not_bob And, honestly, if a site is down for a year, then why bother to keep it in the database?

not_bob They aren't using it.

RN and these options about expiration and bookmarking one would be presented to the user in the consle?

not_bob That's my proposal.

RN oops sorry, hit enter b4 you done

not_bob I was thinking it would be nice if they could auto-expire in the addressbook.

not_bob A simple check, every so often for the leaseset would show that a domain has been active.

eyedeekay I definitely don't want to auto-expire them in the addressbook for exactly the reason you mentioned before

not_bob Understood.

RN I would def want a "pin this one" option

obscuratus Do we have any officical or semi-official existing standard for this?

eyedeekay > i.e. It's good to know that xxx.i2p is the same xxx.i2p you visited three years ago.

not_bob We had a situation about a week and a half ago where someone registered a site, but that site already existed, but had been dead for years.

RN registration is permanent unless the user intervenes as it is now right?

eyedeekay Yeah it's happened before in a particularly suspicious place, too, IIRC a search engine

eyedeekay That's a creepy one to have a bait-and-switch on

not_bob I agree, trust is a major issue here.

RN subscription management tools maybe?

RN where user could select criteria and then preview list and purge

eyedeekay obscuratus, no official standard, we kind of went by what I'm going to coin the "stats.i2p" rule where every domain was permanent because they might come back and squatters could be dangerous

not_bob Yes, that would be nice.

obscuratus At six months, we're coming up quick on the point where zzz.i2p would be released.

not_bob eyedeekay: I understand the logic there.

not_bob obscuratus: Exactly the issue.

RN yes

RN tarbaby right there

not_bob And, we don't what that one to get taken over by a squatter.

eyedeekay Somewhat granular local controls are something that I would be open to

eyedeekay Would it be too difficult to allow the registrant to specify the length of the grace period and save ourselves the decision?

obscuratus On the other hand, the low-lying-fruit would be the people who existing for less than a month and disappeared. Easier to ploink those.

RN that is the thing: subscriptions and address book are responsibility of user

eyedeekay We could get really complicated with it and calculate confidence intervals too

not_bob Very much so.

not_bob Complex.

RN yeah, maybe a registration deadman's switch configrd at registration

eyedeekay OK so it's got to be a fixed interval then ideally

RN like an offline destination will die if you don't tickle it from time to time

not_bob *nod*

eyedeekay And as you say, we're coming up on the time when zzz.i2p would be falling out of the list at the six-month mark

not_bob The other option, which might be better, is any site with less than a month uptime and then six months of downtime gets perged.

eyedeekay But everybody's got zzz's hostname already, we'll squawk at them if a squatter comes down

eyedeekay moreover IIRC it's in our base hosts.txt file still

not_bob Yes

RN I say keep it there

eyedeekay Yeah I basically can't think of a good reason to take stuff out of there

not_bob In the end, it's fine if we say they last forever, domain regs. But, it's also annoying that names get used and then thrown away.

eyedeekay But my point is that arguably when the site goes offline then as long as we don't purge them locally anybody who has a router from when the site was online can tell the difference

RN well the initial reason nb gave... sites that pop up and dissapear

eyedeekay Less than one month of uptime and more than six months of downtime is easy

eyedeekay To approve of

RN but lets say I was offline for 6 mo

RN I would certainly still want to play peek-a-boo

not_bob Agreed.

eyedeekay Well your site has been up for years though

RN mostly yes..

RN think I did have an outage over 6mo in early days though

eyedeekay Could be a year instead?

not_bob I'd fine fine with that.

RN sounds likt this could use more discussion

not_bob Yes, I agree.

not_bob As there would need to be a way to track all this.

not_bob And the router itself my not be the best place to do the metrics.

eyedeekay Yes, but we've been going around in circles about it for as long as I can remember

RN is there a thread about it on ramble yet?

not_bob notbob.i2p/pinger/i2peek-a-boo.i2p.yearly.svg

eyedeekay No but that's a good idea

not_bob Good uptime this last year.

obscuratus Also, isn't each registry essentially able to make their own rules?

RN Σ:Đ

not_bob Yes

eyedeekay Yes, and maybe that's the answer to just have different registries with different rules

not_bob I'm cool with that as well.

RN anyone can make a registry, so yeah own rules. getting adoption for a new one might have friction.

eyedeekay But it may also be true that some rules are best for some situations and some rules are best for others

not_bob There may not be a one size fits all solution here.

eyedeekay But we do need to be thinking about what size fits most

not_bob Thank you for hearing me out.

RN sure, an NSFW blocking one, an anything goes (get your eyebleach), a 'safe' one

not_bob I already do that with my service, offer to subs.

not_bob two subs.

not_bob The non censored hosts.txt is far mor popular.

eyedeekay You're welcome not_bob. Next week, that's what I want us to know is what the possible rules for domain names are and situations where they affect how people navigate I2P

eyedeekay Mostly talking about what we get in terms of security and authenticity from applying these rules, not worried about where it's safe to browse them

eyedeekay Anything else on #1?

eyedeekay #2 netDb Changes in Java I2P

eyedeekay This is mostly just a report on git.idk.i2p/i2p-hackers/i2p.i2p/-/merge_requests/95/diffs

eyedeekay What we've done is found a way to provide Java I2P with "true isolation" of the netDb in a similar way to the way which i2pd has done it by design

eyedeekay But because of the incumbent design of Java I2P, ours has to work a little differently

eyedeekay What we do instead of giving each client it's own netDb to own, we now have a tool which exists to allow clients to access their own netDb as part of a larger, netDb-managing structure using an "isolating key"

eyedeekay That "isolating key" or "dbid" in the code can be anything, but at this time it's the base32 of the hash of the client which received the message

eyedeekay This allows us to keep all the features we had because of our existing monolithic netDb, including extensive ability to graph stats and provide feedback on the network's state, while also providing concrete barriers against attackers who use crafted LeaseSets and RouterInfos to associate clients with the routers that host them

RN [FIREWORKS]

eyedeekay We've been testing it for a while but I would like to encourage anybody who can to build from the i2p.i2p.2.2.1-nested-netDb branch and report their results because this needs to be right

eyedeekay It touches practically every corner of the router, every eye on the code and every tester is very much appreciated

eyedeekay obscuratus thanks again for all your help already

eyedeekay EOT

obscuratus I'll add that the overall framework is working on my testing network, but I am accumulating a back log of fixes that I need to push back.

eyedeekay Let me know about any of them when you're ready

eyedeekay I did check in the excessive local RI publish fix a day or two ago

obscuratus Would it make sense to do MR against idk/i2p.i2p on the i2p.i2p.2.4.0-test1 branch?

obscuratus eyedeekay: Yeah that is now fixed.

eyedeekay obscuratus sure that would be fine

eyedeekay or the i2p.i2p.2.2.1-nested-netdb branch, either should work

eyedeekay test1 is just where I'm merging everything that I'm considering putting onto the release together for testing, I periodically pull i2p.i2p.2.2.1-nested-netdb into it

eyedeekay so I guess prefer i2p.i2p.2.2.1-nested-netdb for the MR

obscuratus Also, for anybody testing it, certain people may notice differences in your router's behavior. Also, the information on the console display can only be described as "mangled" right now.

obscuratus eyedeekay: OK.

eyedeekay Anything else on 2?

eyedeekay Anything else for the meeting?

eyedeekay All right thanks everybody for coming, remember we've rescheduled the next one for Tuesday the 15th

RN :)

obscuratus eyedeekay: For the post-meeting, if you're still around... Some of my stuff isn't nested-netdb related, such as my banlist modifications.

eyedeekay You can put it in an MR?

obscuratus But some of the other stuff is near-by on on-top of the banlist stuff. I'd like to get it in or get it NACKed.

obscuratus Yes, it's ready for a MR, but it's now on to of the nested-netdb stuff.

obscuratus s/on to/on top/

obscuratus Everything change of any size seems to have tentacles going into just enough of everything else to raise conflicts. :)

eyedeekay yeah that's probably part of why zzz really didn't want to do this before

obscuratus I'll put in a MR for the banlist stuff to the i2p.i2p.2.2.1-nested-netdb, and if it looks like you want to somewhere else, just let me know.

eyedeekay Sounds good

eyedeekay It will be fine