The EL backlight adds to the slightly unearthly vibe too. The eMate was one of the last mainstream laptop-size devices made with an EL backlight. I miss those.

This is my view as well. That being said, Time New Roman is marginally better because there are several good, modern open source alternatives with the same metrics that can be substituted. And there's good tool support virtually everywhere for those alternatives, like in TeX.

There is a metric-compatible open alternative to Calibri (Carlito) but it seems more vulnerable to lawyer shenanigans and doesn't have extensive tool support.

Which Times New Roman alternatives would you recommend?

I know a few people who use Quantrix for financial modelling. It is an exceptionally nice piece of software, basically the successor of Lotus Improv, with Improv's more robust and auditable separation of data and formulas.

This is what I recommend too, but for those who want something prepackaged, there's also XigmaNAS, basically a lightweight UI layer and basic configuration on top of FreeBSD. Some of the original FreeNAS developers have been working on the project for almost 20 years.

It's great for people who just want storage and don't want the heavy features that came with TrueNAS' move to Linux (Kubernetes, etc.) or who want full control over vfs_fruit options for serving Macs.

Are you sure it's not called LigmaNAS?

That might be a better name :) The project doesn't have as much visibility as it should because it's gone through several names over the last 20 years: FreeNAS (prior to the TrueNAS split), NAS4free, and now XigmaNAS.

There are a couple recent developments in ZFS dedup that help to partially mitigate the memory issue: fast dedup and the ability to use a special vdev to hold the dedup table if it spills out of RAM.

But yes, there's almost no instance where home users should enable it. Even the traditional 5gb/1tb rule can fall over completely on systems with a lot of small files.

Nice. I was hoping a vdev for the dedup table would come along. I've wanted to use optane for the dedup table and see how it performs.

I think the asterisk there is that the special vdev requires redundancy and becomes a mandatory part of your pool.

Some ZFS discussions suggest that an L2ARC vdev can cache the DDT. Do you know if this is correct?

Yes, that's the reason why a dedup vdev that has lower redundancy than your main pool will fail with "mismatched replication level" unless you use the -f (force) flag.

I'm not sure about whether an L2ARC vdev can offload the DDT, but my guess is no given the built-in logic warning against mismatched replication levels.

Well, the warning makes sense with respect to the dedup vdev since the DDT would actually be stored there. On the other hand, the L2ARC would simply serve as a read cache, similar to the DDT residing in RAM.

One thing to be aware of is that many (all?) current UGREEN boxes can't support ECC RAM. For anyone looking to use ZFS like the article linked here, that may be an issue, depending on one's view of the debate about whether ECC is necessary with ZFS.

Yes, they used to compromise more along these lines in the past. e.g., Samba's vfs_fruit would never have gotten as good as it is without Apple open sourcing their SMBClient. Everyone benefits, even Apple (I'm sure they're running vfs_fruit on their server storage arrays internally). Wish they'd do it more.

My favorite display era was the short-lived period of 4-bit greyscale X-terms. Just enough shades of grey to be usable, few distractions.

There are a lot of greyscale radiology monitors available in the used market these days, but they tend to have a relatively high white point and aren't easy on the eyes.

I was sold on greyscale (2-bit) on the NeXT. I ran many NCD Xterminals back in the day, and the greyscale ones were awfully nice.

I do have a radiology setup I use for writing sometimes, but yeah it's not a general purpose solution. I didn't think about it being the white point so thanks for pointing that out.

I still have nostalgia for my Hercules monochrome adapter with an amber screen.

Agreed.

I have good colour and stereo vision, but I'm very shortsighted. I'm also nearly 60 and now wear varifocals with 3 focal lengths for books and phones, computers, and distance.

But 1980s long-persistence-phosphor CRTs looked good to me, were restful to my eyes, and I could and did look at them all day.

On today's flatscreens I can't see the difference between SD and HD, let alone on a TV across the room where it's imperceptible. I can't tell different refresh rates apart, let alone variable ones. I read excited product releases about tear-free video and fixing things I never was able to perceive in the first place and so which did not need fixing.

Now I am losing my choice of good X11 desktops with rich standards-driven keyboard UIs, replaced by phonelike toy desktops which take 10x the RAM and 100x the CPU and GPU and require a whole new display server, and which can't do nine-tenths of the stuff I used in desktops 15 years ago.

Apparently, because of kids with keen eyes who see things I never could.

I suspect all this stuff about HDR and VRR and gamer screens that refresh at 100s of Hertz are just the same as audiophiles who pay absurd sums for gold-plated 100Mb Ethernet cables.

I want to see double-blinded randomised controlled trials to prove that these people can see what they claim to see -- because I don't believe them -- and to prove to them that the bulk of the population can't tell.

When nVIDIA sells DGX directly they usually still partner with SuperMicro, etc. for deployment and support. It sounds like they're going to be offering those services in-house now, competing with their resellers on that front.

There's a lot of work in this area. See e.g., the LoRP paper by Di et al. There's also a decent amount of work on the other side too, i.e., using LLMs to convert Prolog reasoning chains back into natural language.