Groups are defined and guaranteed correct with the section chain (there is a struct called SectionAuthorityProvider that has the curent agreed elders, section prefix and section key all signed by the section chain)
Nodes need to speak to the current SectionAuthorityProvider nodes to hope to achieve any consensus
Ae is the mechanism to sync this information
How it works is very simple
On speaking to a section include the SectionKey you “think” the group is
on receipt, they will make one of 3 decisions
Key Ok and current - process message
Key is old (we have changed) - return message with ProofChain (the part of the section chain to prove our current SectionKey from that which you gave us) and the SectionAuthorityProvider to make sure you know and get responses from the correct nodes.
Key is unrecognised - return message with an error UnrecodnisedKey
The last part could be you are more advanced than the Elder you are speaking to, but that is cool as we know 100% supermajority of Elders are not lagging.
So all in all Ae allows us to have a SectionActor i.e. a group acting as a single Authority. @JPL was querying this a well recently
I don’t think its realistic to say that SAFE will dominate data or CDN demand any time soon. Obviously, it competes with many other existing protocols in complex overlapping ways and it also competes with direct competitors like ifps, hypercore and others. My most optimistic prediction is that SAFE will be used primarily to store data privately for a growing niche of local-first or client-first apps. The local-first movement is really gaining momentum with the rise of robust CRDT tools. Maybe SAFE is a shoe in there.
Replacing the current set of applications on the internet, IMO, is so far into the future that neither of us will be alive to see even a small dent! Also, SAFE is a high-level opinionated protocol, whereas the internet is full of 1000s of modular low-level standards. Low-level standards are necessary to gain adoption from diverse participates, whereas SAFE makes decisions about currency, encryption, privacy, economics etc…
I sure you’ve thought about this question of low-level vs high-level, but it would be good to understand how SAFE could potentially evolve to encompass a greater level of use cases by being less opinionated.
I am not sure what you mean by opinionated here. The network algorithms/protocols have inputs that define outputs. The client data though is all crdt, so merkle register/multimap and CRDT-tree so far. These are fully open for devs to create anything at all. So the opinionated part here is hard to understand.
In terms of the network consensus protocols they are perhaps opinionated, but I don’t understand the terminology of opinionated algorithms. IMO they all are, but it’s what you can build with the tools that matters.
We have
CRDTs as above
BRB (byzantine reliable broadcast
Fork prevention data (i.e. the Dbc protocol we have to prevent double spend)
Fork resolution protocols (to allow real world concurrent events, so doublespend is required in real world when talking about concurrent data and events)
For non resolving data (app devs can resolve or have users resolve conflict)
Ae
Node aging
For networking quic is used
So in the network, there are a lot of protocols in play. Our issue is we don’t let folk know via docs and posts, but we are on that now.
My big issue (and I mean big, almost red line) is none of these should require advanced maths to understand and they must be simple. I detest complex cryptically named algorithms with pages of greek letters to explain them. To me, that just means we don’t really know yet so here’s a wall of greek that barely anyone understands. There is a niche place for that, like metallurgy to an Engineer but in my opinion, it’s hard to sell a hammer based on a deep technical analysis of the dipole alignment and carbon content. I fear crypto is stuck in that mode, but our approach here is lots of small simple algorithms, each provably correct and easy to understand.
Yes, sorry I should of been clearer. This is primarily what I’m talking about. Using SAFE commits you to a particular economics which is setup to reward early adopters and assumes network growth. In a similar fashion, using Bitcoin commits you to deflationary economics, so the earlier you brought Bitcoin the better.
My main point there is that by being overly prescriptive in token economics SAFE could be held back. If users want a pay once model, then that should be an option, but some users may choose a pay as you go since it will be a cheaper. I could see users skipping SAFE for other options because of this, even if its technology and utility is superior.
It’s important to have the option, but if it’s mandatory then the network is less useful. Not every user wants to store data forever, just like not every user wants a life time subscription to Netflix.
If this is a technical limitation then fair enough, but otherwise I don’t see any advantage to mandatory data perpetuity.
Means web site links never expire, they will always return something. Latest web page if maintained or last known one. With an option to select a previous version. No more broken web.
Personal data can be deleted. There was to be an option to delete it completely, but it is effectively deleted if the person tosses the data map, then no one can read it. Unless you share the datamap.
Also public files need to be forever if people are to be able to read the file any time in the future.
Exactly, it’s more about preserving a historic record of events and making the internet a proper time machine with no broken links/link rot and historical appends than it is about deciding an economic model for everyone.
Not everyone will want their public comments public for forever but there is the option of anonymity or to think harder before speaking.
Data perpetuity a very useful service, but if everyone needs to do it then that will discourage people from using the network, especially when users need to pay a premium for it. Regardless of whether the intention or not is to enforce an economic model, that is the outcome.
There are many reasons why users would want the flexibility of temporary storage for both public and private data. If someone else wants to preserve the data then they can easily pay that additional cost themselves.
In the end anything that discourages use of SAFE makes it less likely that people will use it and more likely they will use an alternative service with more flexible cost and storage options.
I think it’s more of a prescribed philosophical limitation based on what @neo said in his post and it’s the correct one.
The cost for perpetual isn’t going to be much more as storage costs drop - and sure it’s a guess that storage costs will drop, but one most of us here agree on.
Since we think the cost difference isn’t significant, then no need to bother with addition of bulk temporary data - which there are plenty of alternatives for anyway.
There is also the complexity of doing both and having a multi-tiered cost structure that would be confusing for grandma trying to upload her family photo album.
Well for the foreseeable future it will. 100TB drives being sold now. Once the technology matures the price will drop a lot since its silicon. The high cost is in the rejection rate and overstocking the drive so any chips that fail before final test can be bypassed in the firmware and still have 100TB guaranteed.
This technology will filter down (as it is now) and in a couple of years the cost of 20TB SSDs will be a lot cheaper. As I indicated a year ago when inside sources said these 40TB and up drives were being developed, it was because the SSD development went from a niche research & development into main stream research & development. Expect greater and greater developments to appear over he next 5 years, SSD tech is only in its infancy with 3D building of chips to rule in this market.
One area is the development of a standard to make the memory on the drive available to be accessed as main memory albeit slower access speeds in terms of main memory, but allows even faster access of stored data then current M2 drives. Once this hits motherboards and OS then the demand for these drives will boom and the producers of the drives will be gearing up for it.
Storage is leaving the old school ways and evolving during the next decade. And like CPU bang for the buck improvements over the last 50 years expect something like it for SSD/NAND technology over the next 20 years at least.
And other drive technology has risen to the challenge so we may still see more mechanical storage systems provide better cost (but slower access) for quite a number of years.
If we get to holographic storage, then we may jump an order of magnitude or maybe much more. This is probably a decade away at least though … still, things like this are being developed by multiple companies, so who knows.
Yep, storage technologies are being driven by need and profit. There has been work on this for a few years already. I am thinking it will be like SSD/NAND was, slow development till other older storage systems slow down in their rate of increase. These technologies typically follow the “S” curve we also see in adoption of a technology, and there is some relation.
I doubt we are anywhere near the slow down of storage technologies overall, but rather coming out of the slow growth rate and now entering the fast growth rate limited only by demand vs profit. Rotating magnetic disk storage has always had fast and slow growth phases and for decades was almost keeping up with 10 times growth over 5 years average. In one to 2 years we saw SSDs go from less than 10TB to 100TBsimply because they took research&dev from niche to main stream. Imagine the next 5 years will bring as they overcome the 3D silicon frab heat problem and production.
EDIT: And communications is slow to provide faster last mile technologies to the consumers but the backbone fibre is increasing very rapid. The issue for last mile is that the increase to ADSL2+ then Cable 100Mbits/sec plus putting on more people at a high rate and then Mobile increasing rapidly. More people and more people using higher bandwidth means the providers are not wishing to rush last mile development. The backbone is the priority for development, increase the undersea links from 1Gbps to 1Tbps over one to two decades style of increases.
Somewhere on this forum I posted stats for historical cloud data storage (sorry can’t find the link). The difference in annual storage cost over ten years is reliably about 10x cheaper across various time frames and service providers. This works out to be equivalent to about 20% cheaper each year. If we extend that discounting rate to many many years, permanent storage works out to be 5x the annual storage, which is about in line with what you found so that’s pretty cool to see.
I’m not very familiar with the economics of safenetwork token & would like to understand more. Let me apologize in advance if any of my questions below were answered earlier on & you have to repeat yourself.
Is there any paper or site i can read more about this rather than sieve through the many comments.
If I’m a farmer, I earn when data is deposited into my allocated space but not renewal basis, to continue earning meaning I have to continuously increase my space allocation (upgrade storage)?
Is there any incentives to encourage clean up of junk data?
How do we prevent unfortunate event such as government crack down on large farm with alot of data?
Thanks in advance on the expert advice.
Is there any paper or site i can read more about this rather than sieve through the many comments. It’s still a work in progress I believe, but the Primer gives an outline of how it will work.
If I’m a farmer, I earn when data is deposited into my allocated space but not renewal basis, to continue earning meaning I have to continuously increase my space allocation (upgrade storage)? No, you earn when the data is requested by a user, not when it is deposited. You should not need to continually upgrade your storage or other hardware, unlike with bitcoin
Is there any incentives to encourage clean up of junk data? No, because there is no junk data, but I believe they are looking at ways of archiving data that is not requested often.
How do we prevent unfortunate event such as government crack down on large farm with alot of data? Data is replicated around the world, so if one machine or a whole lot of machines get switched off in one particular country it will automatically get copied elsewhere.