robryk @robryk@qoto.org

Weird traffic signs: https://www.google.com/maps/@47.3737534,8.5972604,3a,16.1y,78.64h,88.41t/data=!3m6!1e1!3m4!1sfAwZlrJBrPo2OhH3SHj68g!2e0!7i13312!8i6656

Pe Lang has a temporary exhibition in Paris (in https://www.deniserene.fr/) until ~first week of July: https://bibliogram.art/p/Cdf2MDjseRm

Pe Lang makes artifacts that, for me, show some nonobvious or obvious-but-often-overlooked physical property. https://www.pelang.ch/works.html has some examples (but even not the ones I liked most when I saw his exhibition in, sadly defunct, museum of digital art in Zürich).

It seems that mastodon does server-level blocking ("suspension") of other servers in a way that appears as if it considered all the signatures from that server invalid. Thus, it seems that we can test whether Mastodon server A suspends server B by sending it any public object from B and seeing what the result will be.

That should be non-disruptive, accurate _for Mastodon servers_, and cheap, so it could probably be used to archive the history of suspension graph.

Thoughts on desirability?

To be more precise: the interesting thing is that one can interpret D_JS(A||B) as a mutual information between something for any pair of distributions. (One can do the inverse with D_KL: I(A;B) for any two variables can be interpreted as D_KL between some two distributions.)

TIL (somewhat embarrassingly) that https://en.wikipedia.org/wiki/Jensen%E2%80%93Shannon_divergence provides a connection of sorts between mutual information and KL divergence

Why isn't hyper's Service ensuring the poll_ready invariant with typing?

Docs (https://docs.rs/hyper/0.14.18/hyper/service/trait.Service.html#required-methods) say that the user is supposed to wait until poll_ready returns Ready(Ok()) with calling call. Why not instead have poll_ready return Poll<SomeType, Error> and have `call` be a function on SomeType (that takes it by value and consumes it)?

#rust

When one responds to a boosted post, wouldn't it make sense to by default cc the booster? Is there some obvious reason why that would be unexpected/harmful?

#mastodon

A boost is just an ActivityPub Announce message; there's nothing that forces the message to be public. Why does Mastodon's UI not allow me to boost messages privately?

#activitypub

Possibly trivial: do we know of a group (well, family of groups indexed by the security parameter) where the group operation can be computed in PPT, but you can't invert an element with a nonnegligible chance of success in PPT? What if I also want a PPT algorithm for picking an element of the group uniformly at random?

I can't seem to find an example of such a group. At the same time, I'm pretty convinced that any algorithms that operate on group elements as black boxes can't do inversion using only group operation, comparison for equality, and choose-a-random-element even on cyclic groups (and, I think, even if we also give them an operation that takes group element g and provides a pair (a,b) s.t. a+b=g, a!=e, b!=e that was chosen uniformly at random from all such pairs).

I wonder if we can construct a family of (increasingly large) semigroups of second-preimage resistant one-way functions. More precisely, I would like to have an indexed family (let's denote the index by i) of semigroups of functions (where the semigroup operation is composition) such that:
- i-th semigroup has size >= i,
- the members of the i-th semigroup can be described by natural numbers from a range of size poly(i),
- there are algorithms polynomial in input size that:
a) read i, descriptions of two functions from the i-th semigroup and output the description of their composition,
b) read i, description of a function from i-th semigroup, an input, and evaluate the function,
- the functions in i-th semigroup are resistant to second preimage attacks with security parameter log(i) [^].

tl;dr I want families of efficiently composable hash functions.

#cryptography

[^] I could just as well request a family indexed by i and the security parameter separately (and have all the complexities be poly(input size + security parameter)). This is equivalent.

What's a better programmatic vector graphics library than (py)Cairo? For example, I want to draw a (linear gradient) between two curves, with the color dependent on the vertical position as a fraction of the distance between the curves. It appears that in cairo I need to mess with MeshPatterns to do that, which creates quite a few annoyances (e.g. having to draw one of the curves backwards, have to actually draw the vertical stubs at both ends, etc.).

Are you annoyed by all the weird error handling characteristics of bash? You might wish to take a look at https://skarnet.org/software/execline/index.html or, if you desire compatibility, https://www.oilshell.org/

I would really like for more of my devices to come with this kind of documentation.

This in particular is a multimeter that belongs to my father; it was produced sometime in the late 80s in the Soviet block.

@timorl This is the kind of thing I thought would be happening without eventual consistency in follows :/
QT: https://mstdn.social/@feditips/107916542161347338

I'm somewhat amused that I now own a noticeably loud lightbulb.

@timorl Is this the kind of thing you were looking for when you found mobilizon instead?
QT: https://mastodon.online/@FediFollows/107825507961049716