Recently I've read Vinton Cerf warning us from (1) Data loss due to software updating and not being able to read "old data" any longer and, (2) using URLs to access data, he says a better and more reliable method should be developed.
I'd like to know your opinion about this issues.
Persistence is a matter of management and openness
neuraxon77 1 year 24 weeks 3 days 21 hours ago
XRI and the W3C have been going at it recently over this and a number of other issues. XRI uses persistent i-numbers or URNs and was recommended against by the W3C. See Dave Orchards post for details.
Personally, after thinking about this for a while. I agree with the W3C's stance, while awaiting an XRI members counter argument.
One way I see the cost of managing persistent identifiers being solved, may be a tax(shriek!) but I don't like that idea. How many persistent identifiers would that get me?
Any single point of contact, is a point of failure...
You'd think people and objects really only need one persistent identifier per object that travels with it through it's lifetime. IPv6 addresses comes to mind here. However with this approach, privacy is a real issue to be dealt with.
Any robust system should be self-sustaining...
Peer-to-peer distributed naming is another option. Unfortunately hash collisions mean you move the management elsewhere into crypto systems on untrusted storage and compute nodes. The crypto systems aren't the problem, nor the untrusted storage. It's the compute part performing the crypto and network in between nodes that is the problem. Hence what we have with DNS and centralised management & DNSSEC (which hasn't taken off as I understand it).
I'm currently trying to come up with a better answer - if there is one - other than simple centralised management and paying for a domain every ten years (the current maximum lease period). I think a different peer-to-peer approach could be made to work here. Something I'm currently working on.
The bottom line is that storing anything persistently over generations is never easy.
When it comes to data loss, it's a very complicated problem that leads you down rabbit holes into persisting the hardware that the software uses to transform the data. Open hardware and software & licensing that affords Virtual Machine creation would be a positive step towards this in order to emulate older devices. From memory it's illegal to emulate the ARM CPUs used in most mobile phones, let alone all their embedded peripherals. Mac OSX is also illegal to virtualise as far as I know, let alone the hardware it runs on.
I'm all for openness and letting the community maintain systems where vendors with limited resources come and go along with the services we rely upon.
Not only that but the lessons we can learn from history.
Everything we know now is the result of standing on the knowledge of giants. History is an important teaching tool, it's easy to see how we quickly forget. Past knowledge where possible should be made available for generations that come, in order to relearn from both our mistakes and greatest achievements.
Great ideas are timeless, we should keep them that way!