But in case I wasn't clear when I invoked "Move fast and break things", I'm talking about process, the process by which a body of software like a database is developed, more than the result that people use. Except that I believe both are intimately connected, a bad process is less likely to produce quality software, a sufficiently bad process will not, absent heroic efforts by individuals that tend to be unsustainable.
And taking a step back, into something that's now on my todo list to research, creating a set of rules etc. for a monastic order born in the death throes of the Roman empire, that's endured for 1,500 years and formed a model for many other codes, it not simple, for human nature is anything but simple.
Fundamentally the term "open source" was coined to try to take away the political meaning.
Or because the open source crowd disagreed with the political objectives of the "free software" crowd. We for example define quality differently, and much more conventionally.
If she is a hypocrite then the same CoCs she supports will constrain her.
Who's going to bell the cat? She's already been called out for advocating physically assaulting a journalist who interviewed someone she labels a "Nazi", and it won't take much research to learn that she's incapable of being civil to other people. Read for example her story of her short tenure at GitHub. Or start with her drive by shooting of the Opal project.
Your post might not be "the best information", but it's what I was looking for, also seems also this English version compressed the original quite a bit.
The key would be that "the connotation is 'don't be a smart ass'", which is exactly what we'd expect, being consistent with the rest, vs. something that bans all levity.
There's nothing in the English version that suggests to me the monks would never smile, for example.
Latin may now be a dead language, but surely the meanings and connotations didn't stay entirely static in the long periods before and after this "code of conduct" was written. English... well, outside the Anglosphere isn't it its like Latin in the post-Rome days, a common tongue to communicate, but not sure even back then if it was only for writing.
If/when the Anglosphere goes away, if there remains anything like the Internet (a bit doubtful), we might see something like you're wondering about, where the vulgate would be used in many if not most face to face or real time communications, but words you want many people to understand for a long time get committed to written English. Akin to how English has become the common language for computer languages and systems.
54. Speak no useless words or words that move to laughter.
Emphasis mine. Apparently jokes are right out.:^(
Before getting concerned, I'd want to know what the original Latin meant at the time it was written. For example, if the connotation is laughter at other's expense.... That it's grouped with "no useless words" has to mean something.
However, long-term stability, and keeping everything boring and slow is about the last thing you want in a tech project.
A tech project like a database?
"Move fast and break things" doesn't work for a whole lot of the kinds of data people store in them. (Does actually work for a lot of modern spying on people to sell advertising functions, though, like location data.) GitHub's customers have not been having a boring time in the last little while....
but being over 1000 years old it's amazing how well it still would work as a guideline for a modern CoC.
Rather, the fact that it's stood the test of time for our sorts of communities is exactly why it still works 1500 year later, and an explicit reason for choosing it.
The best though is: "Not to desire to be called holy before one is; but to be holy first, that one may be truly so called", i.e., virtue signalling is forbidden.
History tells us that purity spiraling, in politics, in tech in general, and FOSS as we're discussing, will stop, one way or another, after so many people are killed or banned there's none left to keep it going, or a strongman or men stop it by force, often because they realize they're next. Not exactly sure how this will work out in tech and FOSS, I think there's a good chance we'll split into two groups, all the way down to two parallel Internets, if what's happening in the political arena doesn't also stop it for tech and FOSS.
That all depends on whether or not Moore's Law is truly dead, or only dead for this Intel "10nm" process, with the added wrinkle that non-recurring engineering (NRE) costs have been skyrocketing with these more extreme 193 nm immersion lithography nodes because multipatterning requires many more masks. In the latter case, customers who aren't going to order as many units won't use the newer nodes, unless they achieve some other need of their's, like lower power consumption.
Moore's Law has more been a "law" of economics, in that smaller nodes were for a very long time clearly cheaper to manufacture, since defects were random, especially the classic spot of dust types, and the more die you could fit on a wafer, the higher your yields. Here, Intel has conspicuously failed to get acceptable yields for years, at some point if they don't succeed they will have to pull the plug, either to a less aggressive node, or to the next one if they can get it to work and it's economic. Their "7nm" node is said to use EUV heavily, so the last question is also very much up in the air.
Since yield issues aren't contagious between process generations, we can expect Intel's 7nm to be on track for the originally stated date.
The thesis of the article is that Intel was the first to attempt Self Aligning Quad Patterning (SAQP), in a "10mm" node that's more aggressive than the current or just getting going maybe Samsung and TMSC "7mm" nodes.
Per the author, Intel's "7mm" node will use EUV, which has its own host of problems, but is a fundamentally different technology, see the above link on pushing 193 nm immersion lithography, vs. using what are very close to being soft X-rays at 13.5 nm (per Wikipedia the X-ray range starts at 10nm). Of course, if Intel's 7nm node expected to get some use out of their 10nm SAQP technology, it still might be in trouble.
So the same as any Linux distro, but with OpenBSD I just boot bsd.rd, I choose upgrade, and then pkg_add -Ui. Done.
Well, going by the 6.3->6.4 upgrade, which I just did a trial run of, you may have to fiddle with a few config files, and do some minor tweaking like deleting an obsolete daemon and its user and group, but it's quite straightforward if you have basic command line system administration skills.
I misunderstood your comment to mean you need to do a full reinstall from scratch, meaning configuration and setup. That is not the case.
I'm just now seriously starting my move from Linux to OpenBSD, and just did a trial 6.3->6.4 upgrade of a vanilla only one package installed scratch 6.3 installation. It does look like the process reinstalls all? the standard binary etc. system files, but as you say, the configuration, disk setup, etc. remains the same.
Here's the upgrade guide BTW. It's all manual, but easy for people with sysadmin experience. Replace the bsd.rd ramdisk kernel (for AMD64 it's tiny at 9.6MiB, although I suppose that means it's missing a lot of device drivers Linux and Windows have), tweak config files where syntax or semantics have changed, remove obsolete stuff, boot from that new ramdisk kernel, follow the easy prompts, fix up any mistakes you made in updating the config files (there's a tool to help with that), update your non-base system packages, and you're essentially done.
Not quite for this release, e.g. it turns off Intel Hyperthreading by default, and my OpenBSD scratch system is using a cheap 1 core 2 Hyperthreads Celeron CPU, so I need to turn that back on to test its SMP, and if I was using a sound input, I'd have to carefully turn that back on, it's now disabled by default for the obvious reasons.
So far I'm quite pleased, it's sane like the V7, BSD2.x and 4.x systems I used in the 1980s, but I haven't put it to serious use yet.
Yeah, that's not a use for which the tapes were intended. I still got the impression in the late 1970s from my friends with various audio tape systems attached to their microcomputers that the Commodore PET's was the best of all of them. AlphaMicro also had a VHS videotape based backup system around that time, which my family decided not to mess with. DECtape of that era? Rock solid in my summer's use of one. Magtape in the 1980s, ditto.
Magtape in the 1950s was used as an alternative to cards for primary storage for IBM's business computers until they solved the hard problems of hydrodynamic floating disk heads, they had to do some hard science and engineering using a drum based IBM 650 to crack that necessary problem. Oh, and business customers were much more demanding of reliability than scientific ones in those days, they basically rejected Williams tube based storage, demanding core memory.
Through at least Y2K magtape, maybe not always by then 1/2 inch reel to reel 6250 bpi, was used for primary storage of all sorts of big data sets. Now of course it's just used for backup and archival purposes, LTO tapes write partly overlapping tracks like SMR drives to get maximum density, you can't rewrite a particular tape block or file.
They're still very good at what they do, as long as you observe the environmental requirements (from memory for LTO-4, humidity less than or equal to 50% and "room temperature" for long term storage, you can let them get up to 60% humidity for up to 6 months), and I suppose avoiding buying garbage tapes, that might be a reason some of those 4 companies dropped out of the market. I certainly would never have bought any of their tapes, Fujifilm and Sony all the way.
Tape drive prices are all over the place, there's lots of variables there including of course speed and generation, but you're completely off about media. Amazon prices from a good 3rd party company I've done business with before, Fujifilm tapes, for LTO-6, we're talking $0.0352/GB, LTO-5 weights in at $0.013/GB and LTO-4 at $0.0184.
Newegg prices for 8TB drives, the current sweet spot in capacity, range from $0.0256/GB to $0.0325 for the lower end of the 5 year warranty 550TB bandwidth/year enterprise drives. I hope I don't have to look up per GB prices for the higher capacity drives, companies by them and the higher capacity LTO tapes because space for them isn't cheap.
The LTO drives do cost 10x a hard drive, but you can put cheap media in them, and all day, your claim they are only meant to run a few hundred tapes before breaking, and the tapes are "even much less reliable" is falsified by my own experience with an HP LTO-4 drive and 65 Fujifilm and "HP" tapes over the last 6 years, incremental backups every day, full 3-5 tape backups once a month (and was doing Bacula differentials every week for some years).
Enterprises use them because they're very reliable, less fragile than hard drives which makes them good for off-site storage, etc. Cloud companies, which this article used as a hook since they're currently sexy? Only for archival storage, maybe, like AWS Glacier and Azure Archive, obviously any class of storage that's available within less than a second isn't going to be on tape. See also all the scientific uses where 100s of petabytes which don't have to be frequently accessed are reliably stored on LTO tape (you think they're willing to lose data and thus papers and their careers???).
What is the degradation mechanism on the moving parts of an enclosed hard drive that is not powered? Why wouldn't it work [after 10 years on the shelf]?
Lubricants aren't necessarily designed to survive that long, especially at room temperature. Glues can eventually fail. Heads parked on disk surfaces might build up too much stiction. The plastic in the flexible ribbon cable that connects the fixed electronics to the disk arm might degrade.
Then there's non-moving stuff, flash memory will lose its charge given long enough. Other electronic components might fail, although that's probably a tiny thing, modulo what I don't know about the effects of higher temperatures needed for lead free soldier. The move to that was very badly managed, tin whiskers for example could make bridges.
You're missing something. I've used tape since the summer of 1978, DECtape then, BASF magtape in the 1980s, Sony DDS in the 1990s, Fujifilm and "HP" LTO-4 in this decade, and never had problems. Perhaps you bought tape from one of the companies that just happened to drop out of the market? Sure you stored them properly, and didn't let the drive temp get too high while writing them?
This is the first I've heard of LTO tapes having systemic problems, aside from one machine room incident where the HVAC failed and it was at 120F or more for an extended period.
Are you sure there's any original M-Disc technology in those M-Disc Verbatim labeled BD-XL triple layer discs? The parent company, which made its questionable name in single layer DVD media (inner lot variability was awful), went bankrupt in December 2016, sounds like Chapter 7 where the creditors got all the assets and set up a company named Yours.co to sell discs and such. The first 2 layers have to be somewhat transparent, right?
Sounds like sticky-shed syndrome. Maybe Intel was unlucky in the tape manufacturers they bought from? Or like way too many other companies, were careless about the environmental conditions they were kept?
But you raise a good point, we really don't know if these tapes will still be readable in 30 years. It's just that tape has about the best record in affordable digital longevity. If you want better and also proven, microfilm and fiche might be your best bet, assuming you can still buy it and the equipment for it. Or scrounge old pre-Winchester hard drives, or maybe write to acid free paper, I think you can get 1MiB or more per side that way.
Oh dear, someone else who's knowledge is well out of date. Since LTO4 the tape speed has been variable to avoid shoe shinning.
Which I can most certainly hear as my HP lower end LTO-4 drive reads data off my 15K Cheetah hard drive buffer. But what can the drive do if it doesn't get data for some number of seconds? How low is the lowest rate it will accept data without having to periodically stop the transport?
I think [the LTO consortium members] strive for compatibility two generations back
Yep, normal rule is write one generation back, read two generations back. This only failed during their switch from metal particulate (MP) to barium ferrite (BaFe) as magnetic media, LTO-8 tapes won't read LTO-6, where either MP or BaFe magnetic media was allowed.
DVD-R technology is very marginal, the original pressed DVD version pushed red laser technology as far as it could go. But you still have to buy quality media, for CD-Rs I went with Taiyo Yuden (early on branded as Fujifilm in the US) and not a single one has failed me yet.
For DVD-R I got some from Taiyo Yuden as well, but didn't trust them if for no other reason than that they only cost a cent more than their CD-Rs. MAM-A gold DVD+Rs were my target for "archival quality", ought to test the few that I cut, but they were exposed to bad environmental conditions for a week or more so if they fail that won't tell me anything (note that Taiyo Yuden exited the optical disc business in 2015, selling their stuff to CMC).
Per my readings this month, comsumer BD-Rs are now a disaster, with CMC making the least worst (!!!). No joy there, and as you say, tape just works, still using the HP LTO-4 tape drive I bought in 2011 for ~$1,000 (it doesn't get a fraction of the wear a business would like put it to). Bought 60 tapes of Fujifilm and HP in quantity 2012, 2013 and 2014, plus some odd lots of those brands, and they're still doing fine.
For example Microsoft charges $720/year for 30TB off archive storage, plus a few hundred for bandwidth. I doubt you could do it cheaper with tape, especially when you factor in periodic upgrades, duplication and operating costs.
I think that'll entirely depend on the quantity of data you're storing, and how much you need to access it once it's stored. In terms of raw costs, Azure Archive level blob storage gets down to $0.02/GiB/month for one datacenter no geographic redundancy, whereas as a prosumer I can buy LTO-4 through -7 tapes in quantity 20 for less than that at GB levels (haven't tried converting to GiB).
So I've got a budget of ~$0.22/GiB/year for the overheads you mention, and that's before I have to make any accounting for getting my data back, which at the base price costs $0.087/GiB, plus for Archival storage there's some small retrieval costs as I recall. That cost is fine if I've recovering from a physical disaster while making a 5-6 figure insurance claim, but lots of organizations need more routine, non "archival" access to their datasets.
The real problem with optical is that for the consumer market, there's insufficient demand to keep quality BD-R manufacturers in business, per my research this month, CMC is the least worst manufacturer today (!!!), I'm not sure any BD-R media below CMC's level is going to last even a year.... MAM-A is still making I assume quality CD-R and DVD-R media and today they're the only company I'd trust for those formats. E.g. Taiyo Yuden exited that business and sold their stuff to CMC.
Tape, well, even if Sony is shut out of the US market, there's still the rest of the world for them to sell to. We'll see what happens, I trust Fujifilm somewhat more, but I won't enjoy paying monopoly prices to them. I agree there's no chance the Bay area VC community would even contemplate trying to set up a new LTO tape manufacturing company, but maybe one of the 4 companies that dropped out might restart, if they think they can be competitive while avoiding or licencing, if possible, Fujifilm's patents. Doubt it, though, there's likely good reasons they dropped out.
I'd guess that it's hard to optically read tracks that are spaced at a micrometer apart from each other when all 10000 tracks across the width of the tape need to be read in parallel.
That's not quite how LTO tape drives work, they have tape heads with 8 parallel read/write heads for LTO-1 and -2, 16 for -3 through -6, and 32 for -7 and -8. After writing a "wrap", a single end to end pass, the tape head moves a bit and writes another, partially overlapping the previous set of tracks in the style of SMR hard disks, this allows more density at the cost of requiring a total rewrite if you want to start over.
But your general points stand, except the tape substrate wouldn't have to be clear, you would layer over it a reflective layer, a writeable layer, and then one or more protective layers, with probably one or more buffer layers in-between. But writing at high speeds is a must in this domain, so some sort of multi-track optical head would be required, which strikes me as difficult. Especially fitting it into a 5.25 inch form factor
Looking at Fujifilm's brochures, I see that their metal particle (MP) generations of tapes uses 2 layers over the base, the magnetic material on top, and one between it and the substrate. For barium ferrite (BaFe) two additional layers are added directly to the even more thin tape substrate. And since this is a much more batched process, assume manufacturing tape reels at least 1 foot in diameter, vs. the sort of one at a time method for optical discs, which as far as I know includes a vacuum sputtering step for the reflective metal layer, plus a much thicker substrate, we can imagine why they're much more expensive per GB.
Slashdot Mirror
Maybe....
But in case I wasn't clear when I invoked "Move fast and break things", I'm talking about process, the process by which a body of software like a database is developed, more than the result that people use. Except that I believe both are intimately connected, a bad process is less likely to produce quality software, a sufficiently bad process will not, absent heroic efforts by individuals that tend to be unsustainable.
And taking a step back, into something that's now on my todo list to research, creating a set of rules etc. for a monastic order born in the death throes of the Roman empire, that's endured for 1,500 years and formed a model for many other codes, it not simple, for human nature is anything but simple.
Or because the open source crowd disagreed with the political objectives of the "free software" crowd. We for example define quality differently, and much more conventionally.
Who's going to bell the cat? She's already been called out for advocating physically assaulting a journalist who interviewed someone she labels a "Nazi", and it won't take much research to learn that she's incapable of being civil to other people. Read for example her story of her short tenure at GitHub. Or start with her drive by shooting of the Opal project.
Your post might not be "the best information", but it's what I was looking for, also seems also this English version compressed the original quite a bit.
The key would be that "the connotation is 'don't be a smart ass'", which is exactly what we'd expect, being consistent with the rest, vs. something that bans all levity. There's nothing in the English version that suggests to me the monks would never smile, for example.
Latin may now be a dead language, but surely the meanings and connotations didn't stay entirely static in the long periods before and after this "code of conduct" was written. English ... well, outside the Anglosphere isn't it its like Latin in the post-Rome days, a common tongue to communicate, but not sure even back then if it was only for writing.
If/when the Anglosphere goes away, if there remains anything like the Internet (a bit doubtful), we might see something like you're wondering about, where the vulgate would be used in many if not most face to face or real time communications, but words you want many people to understand for a long time get committed to written English. Akin to how English has become the common language for computer languages and systems.
Thanks a lot!
Before getting concerned, I'd want to know what the original Latin meant at the time it was written. For example, if the connotation is laughter at other's expense.... That it's grouped with "no useless words" has to mean something.
A tech project like a database?
"Move fast and break things" doesn't work for a whole lot of the kinds of data people store in them. (Does actually work for a lot of modern spying on people to sell advertising functions, though, like location data.) GitHub's customers have not been having a boring time in the last little while....
Rather, the fact that it's stood the test of time for our sorts of communities is exactly why it still works 1500 year later, and an explicit reason for choosing it.
History tells us that purity spiraling, in politics, in tech in general, and FOSS as we're discussing, will stop, one way or another, after so many people are killed or banned there's none left to keep it going, or a strongman or men stop it by force, often because they realize they're next. Not exactly sure how this will work out in tech and FOSS, I think there's a good chance we'll split into two groups, all the way down to two parallel Internets, if what's happening in the political arena doesn't also stop it for tech and FOSS.
That all depends on whether or not Moore's Law is truly dead, or only dead for this Intel "10nm" process, with the added wrinkle that non-recurring engineering (NRE) costs have been skyrocketing with these more extreme 193 nm immersion lithography nodes because multipatterning requires many more masks. In the latter case, customers who aren't going to order as many units won't use the newer nodes, unless they achieve some other need of their's, like lower power consumption.
Moore's Law has more been a "law" of economics, in that smaller nodes were for a very long time clearly cheaper to manufacture, since defects were random, especially the classic spot of dust types, and the more die you could fit on a wafer, the higher your yields. Here, Intel has conspicuously failed to get acceptable yields for years, at some point if they don't succeed they will have to pull the plug, either to a less aggressive node, or to the next one if they can get it to work and it's economic. Their "7nm" node is said to use EUV heavily, so the last question is also very much up in the air.
To quote this August article About Intel's 10nm Process Lead:
The thesis of the article is that Intel was the first to attempt Self Aligning Quad Patterning (SAQP), in a "10mm" node that's more aggressive than the current or just getting going maybe Samsung and TMSC "7mm" nodes.
Per the author, Intel's "7mm" node will use EUV, which has its own host of problems, but is a fundamentally different technology, see the above link on pushing 193 nm immersion lithography, vs. using what are very close to being soft X-rays at 13.5 nm (per Wikipedia the X-ray range starts at 10nm). Of course, if Intel's 7nm node expected to get some use out of their 10nm SAQP technology, it still might be in trouble.
Well, going by the 6.3->6.4 upgrade, which I just did a trial run of, you may have to fiddle with a few config files, and do some minor tweaking like deleting an obsolete daemon and its user and group, but it's quite straightforward if you have basic command line system administration skills.
I'm just now seriously starting my move from Linux to OpenBSD, and just did a trial 6.3->6.4 upgrade of a vanilla only one package installed scratch 6.3 installation. It does look like the process reinstalls all? the standard binary etc. system files, but as you say, the configuration, disk setup, etc. remains the same.
Here's the upgrade guide BTW. It's all manual, but easy for people with sysadmin experience. Replace the bsd.rd ramdisk kernel (for AMD64 it's tiny at 9.6MiB, although I suppose that means it's missing a lot of device drivers Linux and Windows have), tweak config files where syntax or semantics have changed, remove obsolete stuff, boot from that new ramdisk kernel, follow the easy prompts, fix up any mistakes you made in updating the config files (there's a tool to help with that), update your non-base system packages, and you're essentially done.
Not quite for this release, e.g. it turns off Intel Hyperthreading by default, and my OpenBSD scratch system is using a cheap 1 core 2 Hyperthreads Celeron CPU, so I need to turn that back on to test its SMP, and if I was using a sound input, I'd have to carefully turn that back on, it's now disabled by default for the obvious reasons.
So far I'm quite pleased, it's sane like the V7, BSD2.x and 4.x systems I used in the 1980s, but I haven't put it to serious use yet.
Correction, for LTO-6 tape we're talking $0.0105/GB, that $0.0352/GB price was for Verbatim single layer BD-R disks in quantity 25.
Yeah, that's not a use for which the tapes were intended. I still got the impression in the late 1970s from my friends with various audio tape systems attached to their microcomputers that the Commodore PET's was the best of all of them. AlphaMicro also had a VHS videotape based backup system around that time, which my family decided not to mess with. DECtape of that era? Rock solid in my summer's use of one. Magtape in the 1980s, ditto.
Magtape in the 1950s was used as an alternative to cards for primary storage for IBM's business computers until they solved the hard problems of hydrodynamic floating disk heads, they had to do some hard science and engineering using a drum based IBM 650 to crack that necessary problem. Oh, and business customers were much more demanding of reliability than scientific ones in those days, they basically rejected Williams tube based storage, demanding core memory.
Through at least Y2K magtape, maybe not always by then 1/2 inch reel to reel 6250 bpi, was used for primary storage of all sorts of big data sets. Now of course it's just used for backup and archival purposes, LTO tapes write partly overlapping tracks like SMR drives to get maximum density, you can't rewrite a particular tape block or file.
They're still very good at what they do, as long as you observe the environmental requirements (from memory for LTO-4, humidity less than or equal to 50% and "room temperature" for long term storage, you can let them get up to 60% humidity for up to 6 months), and I suppose avoiding buying garbage tapes, that might be a reason some of those 4 companies dropped out of the market. I certainly would never have bought any of their tapes, Fujifilm and Sony all the way.
Tape drive prices are all over the place, there's lots of variables there including of course speed and generation, but you're completely off about media. Amazon prices from a good 3rd party company I've done business with before, Fujifilm tapes, for LTO-6, we're talking $0.0352/GB, LTO-5 weights in at $0.013/GB and LTO-4 at $0.0184.
Newegg prices for 8TB drives, the current sweet spot in capacity, range from $0.0256/GB to $0.0325 for the lower end of the 5 year warranty 550TB bandwidth/year enterprise drives. I hope I don't have to look up per GB prices for the higher capacity drives, companies by them and the higher capacity LTO tapes because space for them isn't cheap.
The LTO drives do cost 10x a hard drive, but you can put cheap media in them, and all day, your claim they are only meant to run a few hundred tapes before breaking, and the tapes are "even much less reliable" is falsified by my own experience with an HP LTO-4 drive and 65 Fujifilm and "HP" tapes over the last 6 years, incremental backups every day, full 3-5 tape backups once a month (and was doing Bacula differentials every week for some years).
Enterprises use them because they're very reliable, less fragile than hard drives which makes them good for off-site storage, etc. Cloud companies, which this article used as a hook since they're currently sexy? Only for archival storage, maybe, like AWS Glacier and Azure Archive, obviously any class of storage that's available within less than a second isn't going to be on tape. See also all the scientific uses where 100s of petabytes which don't have to be frequently accessed are reliably stored on LTO tape (you think they're willing to lose data and thus papers and their careers???).
Lubricants aren't necessarily designed to survive that long, especially at room temperature. Glues can eventually fail. Heads parked on disk surfaces might build up too much stiction. The plastic in the flexible ribbon cable that connects the fixed electronics to the disk arm might degrade.
Then there's non-moving stuff, flash memory will lose its charge given long enough. Other electronic components might fail, although that's probably a tiny thing, modulo what I don't know about the effects of higher temperatures needed for lead free soldier. The move to that was very badly managed, tin whiskers for example could make bridges.
You're missing something. I've used tape since the summer of 1978, DECtape then, BASF magtape in the 1980s, Sony DDS in the 1990s, Fujifilm and "HP" LTO-4 in this decade, and never had problems. Perhaps you bought tape from one of the companies that just happened to drop out of the market? Sure you stored them properly, and didn't let the drive temp get too high while writing them?
This is the first I've heard of LTO tapes having systemic problems, aside from one machine room incident where the HVAC failed and it was at 120F or more for an extended period.
Are you sure there's any original M-Disc technology in those M-Disc Verbatim labeled BD-XL triple layer discs? The parent company, which made its questionable name in single layer DVD media (inner lot variability was awful), went bankrupt in December 2016, sounds like Chapter 7 where the creditors got all the assets and set up a company named Yours.co to sell discs and such. The first 2 layers have to be somewhat transparent, right?
Sounds like sticky-shed syndrome. Maybe Intel was unlucky in the tape manufacturers they bought from? Or like way too many other companies, were careless about the environmental conditions they were kept?
But you raise a good point, we really don't know if these tapes will still be readable in 30 years. It's just that tape has about the best record in affordable digital longevity. If you want better and also proven, microfilm and fiche might be your best bet, assuming you can still buy it and the equipment for it. Or scrounge old pre-Winchester hard drives, or maybe write to acid free paper, I think you can get 1MiB or more per side that way.
Which I can most certainly hear as my HP lower end LTO-4 drive reads data off my 15K Cheetah hard drive buffer. But what can the drive do if it doesn't get data for some number of seconds? How low is the lowest rate it will accept data without having to periodically stop the transport?
Yep, normal rule is write one generation back, read two generations back. This only failed during their switch from metal particulate (MP) to barium ferrite (BaFe) as magnetic media, LTO-8 tapes won't read LTO-6, where either MP or BaFe magnetic media was allowed.
DVD-R technology is very marginal, the original pressed DVD version pushed red laser technology as far as it could go. But you still have to buy quality media, for CD-Rs I went with Taiyo Yuden (early on branded as Fujifilm in the US) and not a single one has failed me yet.
For DVD-R I got some from Taiyo Yuden as well, but didn't trust them if for no other reason than that they only cost a cent more than their CD-Rs. MAM-A gold DVD+Rs were my target for "archival quality", ought to test the few that I cut, but they were exposed to bad environmental conditions for a week or more so if they fail that won't tell me anything (note that Taiyo Yuden exited the optical disc business in 2015, selling their stuff to CMC).
Per my readings this month, comsumer BD-Rs are now a disaster, with CMC making the least worst (!!!). No joy there, and as you say, tape just works, still using the HP LTO-4 tape drive I bought in 2011 for ~$1,000 (it doesn't get a fraction of the wear a business would like put it to). Bought 60 tapes of Fujifilm and HP in quantity 2012, 2013 and 2014, plus some odd lots of those brands, and they're still doing fine.
I think that'll entirely depend on the quantity of data you're storing, and how much you need to access it once it's stored. In terms of raw costs, Azure Archive level blob storage gets down to $0.02/GiB/month for one datacenter no geographic redundancy, whereas as a prosumer I can buy LTO-4 through -7 tapes in quantity 20 for less than that at GB levels (haven't tried converting to GiB).
So I've got a budget of ~$0.22/GiB/year for the overheads you mention, and that's before I have to make any accounting for getting my data back, which at the base price costs $0.087/GiB, plus for Archival storage there's some small retrieval costs as I recall. That cost is fine if I've recovering from a physical disaster while making a 5-6 figure insurance claim, but lots of organizations need more routine, non "archival" access to their datasets.
The real problem with optical is that for the consumer market, there's insufficient demand to keep quality BD-R manufacturers in business, per my research this month, CMC is the least worst manufacturer today (!!!), I'm not sure any BD-R media below CMC's level is going to last even a year.... MAM-A is still making I assume quality CD-R and DVD-R media and today they're the only company I'd trust for those formats. E.g. Taiyo Yuden exited that business and sold their stuff to CMC.
Tape, well, even if Sony is shut out of the US market, there's still the rest of the world for them to sell to. We'll see what happens, I trust Fujifilm somewhat more, but I won't enjoy paying monopoly prices to them. I agree there's no chance the Bay area VC community would even contemplate trying to set up a new LTO tape manufacturing company, but maybe one of the 4 companies that dropped out might restart, if they think they can be competitive while avoiding or licencing, if possible, Fujifilm's patents. Doubt it, though, there's likely good reasons they dropped out.
That's not quite how LTO tape drives work, they have tape heads with 8 parallel read/write heads for LTO-1 and -2, 16 for -3 through -6, and 32 for -7 and -8. After writing a "wrap", a single end to end pass, the tape head moves a bit and writes another, partially overlapping the previous set of tracks in the style of SMR hard disks, this allows more density at the cost of requiring a total rewrite if you want to start over.
But your general points stand, except the tape substrate wouldn't have to be clear, you would layer over it a reflective layer, a writeable layer, and then one or more protective layers, with probably one or more buffer layers in-between. But writing at high speeds is a must in this domain, so some sort of multi-track optical head would be required, which strikes me as difficult. Especially fitting it into a 5.25 inch form factor
Looking at Fujifilm's brochures, I see that their metal particle (MP) generations of tapes uses 2 layers over the base, the magnetic material on top, and one between it and the substrate. For barium ferrite (BaFe) two additional layers are added directly to the even more thin tape substrate. And since this is a much more batched process, assume manufacturing tape reels at least 1 foot in diameter, vs. the sort of one at a time method for optical discs, which as far as I know includes a vacuum sputtering step for the reflective metal layer, plus a much thicker substrate, we can imagine why they're much more expensive per GB.