A reminder from someone whoβs been around the block more than a couple times: when you see a PDF or ePub on the Internet that you think you might want to read eventually, go ahead and download it, and squirrel it away for the future. The Internet is not, in fact, forever, and Capitalism will eventually take away anything it canβt rent out to you.
reshared this
Irenes (many)
in reply to mos_8502 π¨π¦ • • •Irenes (many)
in reply to Irenes (many) • • •we're also still very fond of text files.
it used to be quite common for people to write up, say, videogame walkthroughs, or gender transition how-tos, or any sort of information somebody had personally put together, in a single long text file. a quarter of a meg was on the larger end of what you'd see, size-wise (when it's just text, that is a LOT), but not uncommon.
Jonathan Lamothe
in reply to Irenes (many) • •like this
Boyd Stephen Smith Jr. and trisscΜhen like this.
Irenes (many)
in reply to Jonathan Lamothe • • •mos_8502 π¨π¦
in reply to mos_8502 π¨π¦ • • •Shannon Prickett reshared this.
mos_8502 π¨π¦
in reply to mos_8502 π¨π¦ • • •Jonathan Lamothe likes this.
LisPi
in reply to mos_8502 π¨π¦ • • •I recommend people to investigate used enterprise storage.
If it passes badblocks & SMART checking (don't rely on that too much, manufacturers do silly things), it is likely to work just fine, and depending on one's location it is possible to acquire triple-redundant 10TB for less than buying a single 10TB drive new would cost.
mos_8502 π¨π¦
in reply to LisPi • • •LisPi
in reply to mos_8502 π¨π¦ • • •Unfortunately, you're pretty much out of luck.
The used ones are not a good idea (despite looking cheaper), they're often messed up in a number of ways. πΏ
mos_8502 π¨π¦
in reply to LisPi • • •.par2files) partitions, and use good oldtarright to the raw partition block device.LisPi
in reply to mos_8502 π¨π¦ • • •If you ever get to the point of thinking about multi-drive enclosures/cradles though, don't go for USB.
At that point, anything less than SAS (an expander card & breakout cables in an old computer case can do the job quite well) will inevitably prove frustrating through all sorts of annoying issues.
mos_8502 π¨π¦
in reply to LisPi • • •LisPi
in reply to mos_8502 π¨π¦ • • •Conventional Hardware & Software RAID has also been exclusively for stuff where you don't care about integrity ever since 520B sectors have stopped being used. (Those 8 additional bytes were for integrity-checking features.)
Now, such features are implemented instead in modern filesystems since awareness of the data structuring is necessary for sane & reliable recovery (conventional RAID assumes that the drive firmware will both detect *and* notify of errors, which is incredibly optimistic in a setting where general pessimism is the sane attitude).
William D. Jones
in reply to LisPi • • •@lispi314 @brouhaha
1. TIL about 520/528B drives. According to one forum post they might even go back to the 70s ("someone on the Internet said it, it must be true!")
2. Integrity checking features? You mean extra space for higher-level checks besides the (40 or so) extra error correction bits used to protect the 512 bytes of data?
I.e. the wrong data could be correctly written to disk. ECC won't catch that during a read, but the extra 8 bytes could.
πΊπ¦ haxadecimal
in reply to William D. Jones • • •The amount of ECC the drive
adds is unspecified, but even before drives switched to 4K sectors with "512 emulation", the size of ECC was typically much higher than in early 5ΒΌ ESDI/SCSI/ATA drives.
1/
πΊπ¦ haxadecimal
in reply to πΊπ¦ haxadecimal • • •The extra "tag bytes" were mostly intended for file systems that wanted additional metadata associated with the sector, as done e.g. on the Xerox Alto, Apple Lisa and early Macintosh, and IBM AS/400. This could include an identifier for the containing file and offset, but could also be used for filesystem-level error detection.
2/
πΊπ¦ haxadecimal
in reply to πΊπ¦ haxadecimal • • •It is not unknown for drives to erroneously write to the wrong sector, and when the victim sector is read later, the drive will return that _wrong_ content with no error indication. Using the tag in such a way as to detect this kind of error is obviously useful for the filesystem or RAID layer. In the case of RAID redundancy it may be correctable, but otherwise it's still desirable to report an error rather than blindly proceed with invalid data.
3/
πΊπ¦ haxadecimal
in reply to πΊπ¦ haxadecimal • • •4/
πΊπ¦ haxadecimal
in reply to πΊπ¦ haxadecimal • • •To this day, enterprise-grade HDDs (SAS) and even some enterprise SSDs (SAS, NVMe) support reformatting to those slightly larger sector sizes.
I'm not aware of any substantial current usage other than IBM System/I, formerly AS/400.
5/
Jonathan Lamothe
in reply to mos_8502 π¨π¦ • •Jonathan Lamothe
in reply to Jonathan Lamothe • •@mos_8502 :verified: Oh, you already said that...
mos_8502 π¨π¦
2024-09-04 19:52:38