Labo333

This is music notation, not composition!

How are the disk and RAM usage? Compared to elasticsearch and typesense?

It's an information that is typically missing yet very important!

Any idea of how the music was created?

I'm very curious, how do you handle concurrency? Or is this why you have one db per client?

If you really want to browse cat pictures, just use this firefox add-on: https://addons.mozilla.org/fr/firefox/addon/instagram-guest/

There are also proxy websites that allow you to browse instagram without login, but I don't like third parties learning about my browsing.

For such a problem, I really recommend https://kaitai.io/

From their website:

> Kaitai Struct is a declarative language used to describe various binary data structures, laid out in files or in memory: i.e. binary file formats, network stream packet formats, etc.

They even have a Rust interface: https://github.com/kaitai-io/kaitai_struct_rust_runtime

Post an Ask HN :P

Can't wait to get metal support in PyTorch!

Neat! This is actually the fundation of hash-based signatures [1] that are (for the moment believed to be) quantum resistant.

I recommend this awesome series of blog posts to learn more about hash-based signatures [2] (Merkle Trees are discussed in the third article).

Finally, a shameless plug of a repo [3] where I implement Lamport and Winternitz's schemes for one-time signatures and Merkle trees for many-times signatures. I make use of the C reference implementation of SHAKE-128, which is currently considered to be the most secure hash algorithm from NIST.

[1]: https://en.wikipedia.org/wiki/Hash-based_cryptography

[2]: https://cryptoservices.github.io/quantum/2015/12/04/one-time...

[3]: https://github.com/louisabraham/hash-crypto

Unfortunately, the web gets everyday closer to some kind of "interactive TV". Computations are moving to the server side and clients are merely there to get input and display information. An illustration is the very controversial Mighty browser. As a result, programs will shortly become dependent on having an internet connection. Today, people use more Google Docs than Notepad. This removes a lot of freedom because computing now comes at the price of a subscription (for internet but also for other services be it SaaS, data monetization or server costs for self hosting).

A good way to test whether your version is using usleep is

    nm `which sqlite3` | grep usleep

On some servers, it's

    nm -D `which sqlite3` | grep usleep

I think that what the author calls "divide and conquer" is actually "recursion".

Also the complexity of naive fibonacci is exactly the fibonacci sequence so O(2^n) is correct but less precise than O(phi^n)

That's why I use Parallels Desktop, even though I would prefer to use open source software. They provide great file system integration on common OS plus a few other things like GPU graphics (I tried with windows 10 arm).

Thanks but it displays "None" as content.

Any way for me to access the article from EU without subscription or account?

Thanks!

What is your "8C16T Mobile Ryzen laptop"? 12+ battery life sounds really promising!!!

I wish the comparison could also include "message sync across devices". Currently Signal requires one to export past messages in a file (that can be unencrypted). This means that to be sure to keep their messages if they use new devices or lose their old one, one has to use a cloud storage, an area where privacy violating providers are still prevalent.

Note that WhatsApp was doing exactly that (with an opt in IIRC) with google drive sync on android and I guess iCloud on iOS, so I always found even their e2ee claim dubious.

At least Signal doesn't make it easy to use unencrypted cloud storage but I wish to see more services with message persistance. On the other hand, it certainly has a much higher server cost.

Hi, I'm the author of [1]

nodevectors implements many more algorithms than fastnode2vec that focuses on node2vec. I'm pretty sure fastnode2vec is the fastest node2vec implementation because it uses CSR format, JIT compilation but also supports multiprocessing. The sampling algorithm has been improved compared to the paper (and to all other implementations) and allows truly linear memory consumption.

On the other hand, nodevectors has a lot of very cool methods like ProNE so you should definitely try it on your data. However, the original ProNE code is probably faster as it is written in C++ and uses multiple cores while nodevectors just uses the installed BLAS (you can probably get a massive speedup by installing OpenBLAS).

[1]: https://github.com/louisabraham/fastnode2vec

Thanks! I really have trouble to trust in memory databases that require 100% uptime. What happens with a power outage?

Is metadata also replicated in S3?

Else I don't understand how the metadata can be persistent after reboot as AFAIK redis cannot dump and reload its state.

Discord is awesome but does not have support for end-to-end encryption AFAIK

UPDATE: it looks like m.facebook.com now shows the same as mbasic.facebook.com. They are probably reorganizing their websites.

> mbasic has been unmaintained for a while, in my experience.

true

> Lately, many links (e.g. group posts) point to m.facebook, which means that I just get redirected to the home page.

Conversely, in a browser without js, m.facebook.com redirects to mbasic

> You may need to access the page by clicking a link from another mbasic page in order to have some kind of checksum in the URL.

Good idea! From the main page with the feed, I can see the number of new messages but when I click on the link, I see an error message (same as with direct links).

Yes, I get the same! I was using mbasic on mobile and messenger.com on my desktop. But an update last week made messenger.com very slow on my ARM laptop.

I hope it's just a coincidence that messenger.com is updated and mbasic stops working for messages during the same week!

I might have an explanation for the reduced RAM consumption thanks to the M1 chip: RAM compression (https://www.lifewire.com/understanding-compressed-memory-os-...).

I'm using zram on manjaro and I see it as a trade-off between RAM and CPU power.

As specified in the post: "100% free"

I have used the Pinebook Pro as a main driver for nearly 6 months because my main died before isolation.

> Alt-tabbing between Firefox and a terminal takes one second, as does switching between Firefox tabs.

Using the OpenGL compositor (which should be the default in latests versions of Manjaro) is a lifesaver! Web apps are and will still be slow sometimes (I don't use slack but messenger can be sluggish).

> Video playback on browsers is not really nice.

Depends, YouTube is indeed quite slow (I think it became slower after some YouTube update), but if YouTube was optimized for speed there wouldn't be all those adds, right? I like to listen to YouTube music in the background, I use https://github.com/mps-youtube/mps-youtube for that and it works like a charm, uses 10% CPU and a few MB of RAM. I don't use mpsyt for video but I guess it would also improve your experience. Other websites (kissanime) work like a charm, even in 1080p.

> Manjaro does not provide a way to disable tap-to-click True, but installing `xf86-input-synaptics` (although it's deprecated) provides such an option. Also, I use ctrl+F7 to disable the touchpad while I type large chunks of text.

From the specs, one could think that the RAM is the problem but zram totally solves that (I have dozens of tabs open in Firefox, including heavy apps like Messenger, Trello and Overleaf). The slow CPU is a larger issue, with Gmail being so slow to load it's almost unusable (loading my mailbox takes 20s and creating a new email takes 3s).

Overall, I find the machine very much usable though. I only have two complaints:

- static noises

- under "heavy" load (eg html5 games with compilation in the background), the alimentation is less powerful than what the machine consumes, which leads to the battery discharging.

I wrote a guide on my setup that includes other tricks. Shameless plug: https://louisabraham.github.io/articles/pinebook-pro-setup.h...

Now, posts full of emojis that use gradient descent for a classical problem with an efficient analytical solution (https://en.m.wikipedia.org/wiki/Linear_least_squares) that you learn of in any ML class make it to front page...

I think they use normal PN, not RPN.