itnuts

A separate thread is spawned to ingest data into the pool, but it doesn’t stop to wait until the tasks are processed. Therefore, if task definitions consume substantial amount of memory it can lead to OOM kill.

If you’re using multiprocessing.Pool, consider upgrading to concurrent.futures.process.ProcessPoolExecutor, because it handles killed workers better. It doesn’t affect the problem described in this question.

https://stackoverflow.com/questions/40922526/memory-usage-steadily-growing-for-multiprocessing-pool-imap-unordered

a quiz question would be to expand the abbreviation.

list of timezone names: https://www.timeanddate.com/time/zones/

When saving an image from the graph-driver store, a new archive is created containing the uncompressed layers; saving/loading will produce the same layers as were pulled, but the save/load won’t include information about their compressed digests, because this information cannot be verified without the actual compressed artifacts (as pulled from the registry). Reconstructing the compressed layers is not possible due to compression algorithms not being 100% reproducible (they may be most of the time, but various factors, including CPU load, and CPU optimizations during compression can cause their checksum to differ).

https://github.com/moby/moby/issues/46664

Price per 2TB

• GDrive $137.75 / year
• Dropbox $120 / year

Privacy

When I shared a file with someone, and then want to: unshare and share with someone else.

Can the sharing link be reset for a particular file / folder?

• GDrive: copy file, then link will change. No way to reset the link. community answer from 2021
• Dropbox: you can reset, set expiration date, etc dropbox docs

NOTE on Dropbox:

You can only delete shared links that you create. If you invite someone to a file or folder with editing access, they can create their own link to share with someone else. You can’t manage those shared links

ref: https://toot.cat/@plexus/112403882840794498

https://arstechnica.com/security/2024/05/novel-attack-against-virtually-all-vpn-apps-neuters-their-entire-purpose/

general practice for VPN, is to prioritize itself via defining more specific routing rules, which take priority over the default gateway. the attacker, however, can spin up another DHCP server, and use rule 121 to push some new routes to the client. Importantly, if the routes are more specific than the ones defined by the VPN, they will take over, and unencrypted traffic directed to the VPN interface will end up on the attacker’s machine.

the golden standard to avoid this vulnerability is using network namespaces: https://www.wireguard.com/netns/#the-new-namespace-solution

discussion: HN1 , HN2

https://en.wikipedia.org/wiki/Gartner_hype_cycle

Terry Pratchett, «Equal Rites»

I discovered that in fish, process substitution <() works via temporary files, and >() is not supported at all (maybe because it looks like fish?)

Relevant discussion since 2014, and still active now: https://github.com/fish-shell/fish-shell/issues/1786

fish is black of shells

i.e. authoritative and with good defaults.

Didn’t try it myself, but it is very tempting with its command completions inferred from man pages.