chellomere

joined 2 years ago
[–] chellomere@lemmy.world 2 points 3 days ago

You should get some, they are rad

[–] chellomere@lemmy.world 1 points 5 days ago

I have one that I bought in 2013. Has fallen to the floor countless times, still works like day 1.

[–] chellomere@lemmy.world 3 points 6 days ago

It is a killer combo!

[–] chellomere@lemmy.world 8 points 6 days ago (2 children)

I'm happily running an Intel NUC as TV computer since 2013, and it's awesome for exactly the reasons you state. I invested in it when I realized how fully crap the "smart" features of my Samsung TV are. The ultimate controller for it is a combo keyboard and touchpad, I have the Logitech K400r.

The NUC is starting to show it's age now with its 4th gen i5, and I'm in the process of replacing it with a mini PC with an Intel N100.

[–] chellomere@lemmy.world 3 points 1 week ago* (last edited 1 week ago)

It would make sense that a code editor could use a more limited subset of text rendering that could be more optimized.

Perhaps a bit surprisingly, Microsoft actually has pretty good documentation on OpenType. Here's info on what shaping applies to "standard" scripts:

https://learn.microsoft.com/en-us/typography/script-development/standard

And here's the landing page for the latest OpenType spec:

https://learn.microsoft.com/en-us/typography/opentype/spec/

Specifically for ligatures, you're looking for the liga feature which is specified in the font's GSUB table.

[–] chellomere@lemmy.world 10 points 1 week ago* (last edited 1 week ago) (3 children)

HarfBuzz does opentype shaping, that is, transforming strings of unicode characters to lists of glyphs with positioning. The significance of this can be hard to understand for someone used to the latin script, as that needs very little shaping - kerning is often the only thing that's absolutely necessary.

But in complex scripts, most notably the Indic, there's a lot going on. Unicode characters can merge into one glyph under circumstances, or one character can split into several, and relative positioning in both the x and y axis is imperative.

A reason that OpenType shaping is complex is that part of the rules for what to do will be found in the font, and part will need to be hard-coded in the code implementing it.

If you're going to roll your own text renderer, you'll have to care about the following areas:

  • Rasterization/rendering to bitmaps, including hinting (notoriously difficult, old-style TrueType hinting instructions are bytecode, so you'll be writing a tiny VM for this)
  • Shaping (Kerning at a minimum, full OpenType shaping for international support)
  • BiDi (for full international support, primarily Hebrew and Perso-Arabic)
  • A caching system for rendered text glyphs and shaped text runa, as it will be too slow to perform this each time you want to render some text

Let's just say that I do not recommend going this route unless you're prepared to spend a lot of time on it.

[–] chellomere@lemmy.world 3 points 1 week ago (1 children)

Y'know, you can turn off auto correction. It's the first thing I do with any keyboard on any smartphone.

[–] chellomere@lemmy.world 3 points 1 week ago

Hey man, there's worse things to be than point of sales!

[–] chellomere@lemmy.world 8 points 1 week ago (1 children)

What kind of socks do you need for Linux Mint? Are any mint green ok?

[–] chellomere@lemmy.world 3 points 2 weeks ago

Cool, I'll just whip out my peptide printer and try it out!

43
submitted 5 months ago* (last edited 5 months ago) by chellomere@lemmy.world to c/selfhosted@lemmy.world
 

So, I currently have a Netgear ReadyNAS 314 with 1 SSD, 3 HDDs, Intel Atom D2701 and 4GB RAM, running Debian 12, and since getting it I've been getting more into self hosting. What I have now is primarily too weak in the CPU and RAM department, but it could also use more HDDs. I'm aiming for 5-6 3.5 HDDs, 1 Nvme, 1 2.5" SSD.

What I'm currently running:

  • Samba and NFS server

  • OpenVPN

  • Jellyseerr/Jellyfin/*arr stack

  • Pangolin

  • Dawarich

  • Immich

  • rsnapshot

  • Homepage

And it's rather sluggish right now, and is almost filling up its 4GB of swap.

What I'd also like to be able to run/have:

  • Nextcloud

  • Transcoding (including ability to decode AV1, but preferably also encode)

  • Anything else I may want to run (working on degoogling myself)

  • ECC RAM (to prevent bitrot, I'm already running btrfs raid1 to prevent bitrot from faulty disks)

  • 1x 2.5G ethernet

If possible I'd like to have some room for upgradeability. I'm aiming for a low power build, that should be rather compact, especially not very wide unless I can find a better place in my office for it.

I'm looking at a Jonsbo N1 chassis (17cm wide) , but I'm also following a Readynas 626 (19cm wide) in an online auction. Options:

Intel N100 board

Pros: cheap, low power, quicksync with av1 decode

Cons: boards with 2.5G ethernet have to be ordered from Aliexpress and have no support and uses the JMB585 chip that prevents low power C states, limited pcie lanes, no AV1 encode, not very upgradeable (1 DIMM, soldered CPU) , no ECC, I worry it may be too slow

Intel 13100

Pros: AV1 decode, quite fast, upgradeable

Cons: No ECC, relatively expensive, no AV1 encode

AMD 8500G

Pros: AV1 enc/dec, ECC, relatively fast, upgradeable

Cons: relatively expensive, not as low power as the 13100

Readynas 626

Pros: enterprise grade HW, less DIY, ECC, may be relatively cheap

Cons: high power for its performance (roughly that of the N100), wider (19cm) than a Jonsbo N1 (17cm), not upgradeable (no CPU or mobo swap), expensive DDR4 2133 ECC UDIMM, doesn't have M.2 but has a PCIE slot

I'd love to hear what you think about these options and whether you have other concerns that I haven't thought about.

Edit: I just now realized that the 13100 doesn't have AV1 encode in HW, that didn't come until Core Ultra. And wowee, suitable mITX mobos start at 400$ here! I think AMD is the realistic choice if I want to go for AV1 HW encode...

view more: next ›