You actually can verify the piece. The storage node stores a signed piece hash side by side with the piece.
The storagenode can, but the satellite canāt and also canāt trust the storagenodeās claim.
It is a signed piece hash. The storage node canāt manipulate it without breaking the signature.
The repair job is downloading just 29 pieces and uses the signed piece hash to identify corrupted pieces. So the satellite is using that information today.
I wasnāt aware of that (and didnāt notice the word signed in your previous post). I could have sworn I read somewhere that the repair/audit process used the RS decoding process to know which pieces were corrupt. Perhaps this is true for audits, as it doesnāt receive the entire piece, so the hash wonāt work. However, This doesnāt solve the issue that there is no affordable option to find out which of all pieces of a node are lost/corrupt. So itās not a way around DQāing nodes which lost more than 4% of data.
Originally that was the case so you might have seen it in some old blueprints or so. With a test case I demonstrated that it requires 30 pieces to identify corrupted pieces with RS. The audit job has enough pieces to use that method but the repair job was build to download just 29 pieces. In my test I damaged one of the pieces. The repair job didnāt notice it and ended up uploading a high number of bad pieces to the point that the file was corrupted. ā Code was changed to pass that test. The signed piece hash was born.
Thank you I didnāt want to argue with people directly.
CP
Friends, someone, please post a transcript of the event
Audio will be available here https://twitter.com/i/spaces/1djGXlAqDOVGZ
Also note, weāre starting 30 minutes earlier for allow for more questions, but will still be available during the original advertised time for those who can join.
Hi All, Bryan here from the Ops team at Storj. Thank you all for your questions for our upcoming twitter space. We hope to address many of these in the live (and recorded for later consumption) twitter space but we will also plan to respond to many of these questions afterwards here in the forum. Thanks again for the input and contributions looking forward to continuing the discussion.
The Twitter Space is live now! Please join us!
Cant talk or post in the twitter.
If you were to remove the /24 limit and implimented a deposit factor, would the held funds be waived?
if the edge service is so expensive, why you not decentrolize it in first step to lower this cost, and give SNO benefit from it?
Given the expense of edge services. Have you considered differentiating pricing, perhaps by lowering cost of native integrations and adding a small additional charge to use edge services?
This adds an additional incentive to use native or self hosted S3 gateway.
Did you happen to address the native backup applications?
mey be storj can sell some self hosted plug and play mashines to client that they can host in their office and use it like dropbox
Maybe as a round up, what is your initial interpretation of SNO responses and has it changed the way you look at the proposed payouts? Or maybe made you shift or narrow the possible payout ranges for the future?
Not sure what went wrong with my audio there, but unfortunately twitter didnāt want me to become a speaker. Thanks for the feedback and answering the questions as posted on the forum.
It seems twitter has been very flaky with these spaces. Iāll try again next time though.
i dont see at all possibilities event to ask to speak on the web, discord would be better platform.
Someone, donāt remember who, said the request to speak button is in the phone app only.
I listened on Chrome on desktop and did not have a request to speak button so this might be correct.
Thereās nothing currently in Twitter Spaces documentation regarding this issue
Yeah, Iāve only ever seen that option in the app.