3rd Community Review of NonEntropy Allocator

[joshua-ne]

Allocator Application: filecoin-project/notary-governance#1022
First Community Diligence: #63
Second Community Diligence: #148
Allocator Report: https://compliance.allocator.tech/report/f03018029/1731979795/report.md

Since the last community diligence and we got a refresh of 5 PiB Datacap, the following are the updates:

1. New Client
   We have served two more clients: [DataCap Application] <Xenogic> - <Xenogic027> joshua-ne/FIL_DC_Allocator_1022#34, [DataCap Application] <Byte Tunneling> - <ByteTunneling_data_store_bc_fil_01> joshua-ne/FIL_DC_Allocator_1022#39, and [DataCap Application] Pangeo Community joshua-ne/FIL_DC_Allocator_1022#41

2. Progressive Allocation
   We continue to manage the Clients' application and approval process in accordance with the guidelines of DataCap governance. We always start with a small amount, building trust by checking data distribution and SP retrieval rate. For some of the suspicious behavior, we adjusted a little bit, for example, with [DataCap Application] <Byte Tunneling> - <ByteTunneling_data_store_bc_fil_01> joshua-ne/FIL_DC_Allocator_1022#39, we allocated two consecutive 50TiB to begin with, since they did not update their change of SPs on time.

3. Data Distribution
   Our Clients allocate DataCapto different SPs in reasonable proportions and ensure that SPs do not store duplicate CIDs.

4. Emphasis on Data Retrieval
   We have been using both SPARK and our own retrieval tool, which is now OPEN SOURCE (and we have introduced this new efficient tool to the community at a meeting, https://www.youtube.com/watch?v=XQlyGV4N_y8, starting 44:10) and which we believe is more suitable for allocators' usage. For example, we have seen SPs taking advantage of the SPARK system by selectively broadcasting deals, say only 10% or even 1% of them, but still get a good retrieval rate. However, this method will not pass our checking, since we check on the deals made between specific clients and SPs with given DataCids of specific batches. We always check before we approve, and we have rejected several signings due to the fact that they do not perform well on either SPARK or our own retrieval system.

5. Dataset Card Initiative to Make Data More Valuable
   To help the community better use the data onboarded on Fil+, we have started a new initiative called Dataset Card, [Dataset_Card]: <Sample Dataset> joshua-ne/FIL_DC_Allocator_1022_Dataset_Card#1. We are testing and practicing on our past and ongoing clients. With this mechanism established, the community will have much better access to the real data on Filecoin which may lead to easier development on Filecoin.

As we reach out to more clients and SPs, we understand more about their needs and challenges in onboarding data to Filecoin network. From what we've seen, the majority of them are willing to store and keep the unsealed copy of the deal data, it is just we need to do more work to make the onboarding and retrieval easier.

As our operations expand, we are collaborating with an increasing number of clients, which has led to a faster utilization of quotas and shorter application cycles. Therefore, we hope to secure a higher quota in this round. We deeply value the principles of FIL+ and have consistently been actively involved in the FIL+ community. We have contributed tools such as retrieval platforms and CID databases, making these mature tools open-source for the community. With this commitment, we will continue to handle and utilize the granted quota responsibly and prudently.