§HotNets
Organised travel dates, travel, and accommodation for HotNets.
I was awarded a Queens’ College Perry Travel Award to the value of £740!
§Met with Anil
Agreed to schedule dates to visit NYC and then find a time to meet Jane Street to discuss the polyglot package management project proposal. I’m going to take the train from Boston to NYC! Anil said I’ve got to get a Junior Ranger Badge from a National Park in NYC.
Anil recommended live blogging the conference on Mastodon.
Talked about what a HotNets presentation could look like, and whether a live demo was feasible. Maybe I could pull AEON and Patrick’s OCaml R-Tree together into something workable. I think while very cool the APIs just aren’t there for the AR demo on a hololens. A smartphone (e.g. iPhone LIDAR) could be a lot more feasible though… I think considering the goals here is a useful exercise. I’m going to say the goals of this presentation are:
- To get people excited enough to read the paper.
- To get people to come and talk to me about ideas for or issues they see with the SNS.
This should inform a SIGCOMM paper.
Anil asked if AEON is production-ready, and I said I’ve been running it on my server for (according to my NixOS config git history) 9 months! There is a lot of stuff around tunnelling in there that complicates things, but that is useful for the DNS over DNS Oblivously (DoDO) idea. We’ve also discussed a signpost-like idea that AEON could be adapted to taking on similar functionalities to a TailScale control server. I think Jane Street has a Noise protocol WireGuard implementation lying around that could prove useful.
Anil cautioned me against spending time on Hibernia. He said the idea of a service graph for things trying to run that could be dynamically pushed into machines and migrated is more interesting. We’ve discussed a similar idea realting to Eilean before.
Anil said when writing or presenting to consider the input mental state of the recpients and the desired output state. Every paragraph is a process, control extraneous thoughts.
Anil encouraged keeping multiple projects in a ‘flywheel’.
We discussed deploying Matrix servers on eboards on physical spaces for the EEG.
§Met with Anil and Magnus
Discussed the R-Tree based ARP (a link-layer independant spatial resolution).
Similar to ray-casting game engines, responding with RRs.
Eval - 2 scenarios:
- low latency
- global & local
Magnus mentioned frequency scaling.
Anil encourage thinking about the full connection pipeline lifecycle (in <20ms).
Magnus is thinking a lot about how to bootstrap this, e.g. how to get the DNS server you resolve to with 5g… or just your physcial presense without any network connectvity.
PUB/SUB came up a few times, but I think request/response is more approriate. You can return multiple responses to a query. Maybe PUB/SUB would be useful for mobile devices.
We talked about access control, getting quite fine-grained (guests in your house not being able to turn your themostat on). I think some of this should be in the pervue of the naming system (e.g. can you address this device), but some should be left to the other policies and mechanisms otherwise it gets way too complicated too quickly (expressing access control over functionality of a device). DNS update keys could be one way to authenticate access control.
Talked about iPhone LIDAR’s. How you could scan to create zonefiles (automatically ‘statically survey’), translating 3d space into a 1d network topology.
Magnus is thinking about the load placed on these services especially for geodetic resolution.
We think the local low-latency resolution might be the best use case for the SIGCOMM paper.
On EIO Anil said it’s not worth making this a unikernel, just to use the Mirage libraries.
Anil said Magnus might want to try deploying AEON as a next step.
§Shark
§SRG Seminars
Organised SRG seminars for Noa Zilberman’s students to give practice talks. Advertised the Xen SRG seminar.
§Website
Added disseration(s) to papers section.
My patchwork of makefiles, bash scripts, pandoc invocations, and more is starting to get a bit unwieldy. A good project for a rainy day would be to adapt Patick’s OCaml Static Site Generator.
§Matrix Synapse authentication with Raven Shibboleth
Got a good resource at https://nickx.hu/posts/2020-07-13-matrix-synapse-shibboleth-saml.html Spun up an experiment on a Matrix server I had lying around on an Eilean test instance.
§Teaching
- TA’d the last L50 lab.
- Marked and supervised Distributed & Current Systems for Pembroke.