Does zfs have any special 8.3 file name features, like apparently ntfs?

myrkraverk_ um, only special name popping in my head is .zfs in root of the dataset

Oh, I was glancing through devblogs.microsoft.com/oldnewthing/20050715-14/?p=34923

and wondering now if ZFS ever had a similar feature, since I know it was meant to share to different operating systems.

Maybe I guess it's always just been part of Samba, and never Unix file systems?

the short name support is more about SMB property, but zfs itself is having casesensitivity, normalization, and utf8only properties affecting the names; I can not remember if illumos zfs has utf8only or not, I think, it does

Is something like this possible with an extension to ZFS proper?

If I tried to write said extension, that is.

Or would that have to be submitted as a patch to zfs?

finally, my lan is connected to internet again:)

tsoome_, internet connections are so overrated.

yea, and those new appliances all need internet and cloud even to start to work or at least to enable access to all features:/

yep. Our new dishwasher's "delay start" is app-only. And my dishwasher's ipv6 stack is wonky. I've observed it do a neighbor solicitation for a DNS recursive nameserver that's not on-link.

@sommerfeld ==> Brand? I bought my dishwasher a couple of years ago when its wifi stack was an add-on extra we refused.

We have a new washer/dryer combo that might have app-only features but we don't apparently need 'em as the front panel DTRT.

(pardon latency... meetings start in 1min)

danmcd: Bosch - one of the higher-end models.

we also have an LG washer that's wifi connected. The wash-completed notifications are useful (on both appliances).

tsoome: sorry it took a bit to get to, but your efiserial changes appear to work with azure

jbk great, thanks!

the completion notices are really useful, unfortunately jura does not believe in those, so I really have to be watchful to save my coffe from receiving extra water from milk system wash cycle... :P

i wonder if it might be worth throwing up the hyper-v bits to upstream now...

definitely:)

with the understanding that there's no documentation, so a lot of answers are likely 'because that's how the support on FreeBSD works' or 'i don't know' :)

btw, have you tested 16606 with non-ssds?

jbk: Is this basically the original delphix work + your stuff for v2?

tsoome: yes, are you having issues w/ it?

There were definitly a lot of things that weren't just 'because FreeBSD' in the last review effort.

rmustacc: pretty much, though I tried to improve the vmbus stuff as well so it was at least a bit better as a nexus bus

no, I have been busy with getting access to my heating etc systems but I intend to create build with it and see:) I happen to have 4 SATA disks in this machine:)

jbk: Gotcha. Well, I'm hoping a bunch of the reivew feedback from last time didn't just get thrown out. :/

no, i remember going through and trying to address what I could, though it's been a long time since that bit was done

All good.

if andyf still has that somewhere, we could go back over it with all of the changes

I think just the caveat is that even if the answer is because freebsd, we'll just want to understand why where practical. I realize it's not always practical.

as a 'first' pass

tsoome: my home server has 4 sata drives as well, so i tested it there too and at least with those disks, all looked good

good.

ACS-3 said the SATA log pages it uses for that info should return 0s if not supported, unfortunately it doesn't really say if any are mandatory or not for certain things

which given SAT-5 says most of the time such scenarios are undefined, and the logs pages themselves don't appear to report anything too exotic

i'm working on the assumption they're probably fairly standard and not worth the trouble of worrying about disks where those pages might not exist unless examples can actually be found to necessitate the additional complexity

since the failback if they don't right now is you basically get 'empty' values or such

I would probably worry about the more likely case of say qemu not implementing it.

yeah, virt is the worry there

because people do understandably implement as little as necessary, often.