BIT-0016: Subnet Deregistration

[gztensor]

Summary

BIT-0016 proposes reenabling and improving subnet deregistration to remove non-functional subnets that waste emissions, chain storage, and compute resources. Deregistration would be based on on-chain metrics such as lack of demand for the subnet’s Alpha token, low price and emissions, centralized miner distribution, and poor liquidity/volume parameters. A new SubnetLimit (default 128/256) would trigger automatic pruning by removing the lowest-emission subnet outside its immunity period, with tie-breaks by oldest registration. Deregistration would return the registration cost to the owner (or owners in case if it was a crowdloan), fully clean up α-in/out stakes, distribute remaining TAO to stakers, and adjust owner refunds based on emissions received, with root-only access for dissolution calls.

[mcjkula]

I assume that's outdated, because it still mentions the old behavior (lowest emission)?

[gztensor]

Yes, that's just whatever starting point design was. It's an open discussion.

[mcjkula]

Suggested change

	```mermaid
	```mermaid
	flowchart TD
	A[New Registration] --> B{Check Subnet Limit?}
	B -->|No| C[Register New Subnet]
	B -->|Yes| E[Prune a Subnet]
	E --> F[Deregister]
	F --> G[Register New Subnet]
	H[Deregistration<br/>] --> I[Destroy α-in/out]
	I --> J[Distribute TAO]
	J --> K[Remove Network]

[mcjkula]

It has both the mermaid chart and the old content in the code block right now (syntax broken)

[nikola-rahman]

Would be great if there was a plot showing how current subnets score for each of these criteria.

[mcjkula]

Since the DTAO launch, the non-functional subnets keep consuming emissions, as well as chain footprint and compute resources.

Wouldn't that still be the case by setting the SubnetLimit to 256, given that 128 subnets have to register first, until those currently non-functional subnets would get deregistered? Is the focus then, that while they would continue to consume emissions, at time of their possible deregistration in the future, the TAO would be "freed" again for redistribution for different subnets?

[mcjkula]

Only allow deregistration of subnets which would have a higher value of Alpha from liquidation than from unstaking

The table below shows which subnets would be the "first ones" to get deregistered should they be not in immunity based on the criteria of having a higher value of ALPHA from liquidation than from unstaking, sorted/ordered by lowest price first.

Netuid	ALPHA price	NAV Ratio
126	0.001	1.3278
103	0.0011	1.4477
105	0.0011	1.0102
128	0.0013	1.4157
122	0.0013	1.3143
86	0.0013	1.2147
102	0.0013	1.8806
127	0.0014	1.6807
109	0.0015	1.7133
94	0.0015	1.3397
111	0.0016	1.7108
107	0.0016	1.7597
108	0.0017	1.3951
97	0.0017	1.1824
121	0.0017	1.9053
74	0.0017	1.0838
90	0.0017	1.9745
59	0.0017	1.1223
66	0.0018	1.1114
38	0.0018	1.2223
80	0.0018	1.0553
60	0.0018	1.3497
24	0.0018	1.3141
47	0.0018	1.4738
70	0.0019	1.4593
104	0.0019	1.2621
91	0.0019	1.2304
76	0.0019	1.4825
113	0.002	1.0707
79	0.0021	1.5682
114	0.0021	1.7816
28	0.0021	1.2822
27	0.0022	2.1783
58	0.0022	1.6837
26	0.0023	1.177
55	0.0023	1.6008
99	0.0024	1.8332
21	0.0024	1.3293
84	0.0024	1.1841
46	0.0025	1.406
61	0.0026	1.4416
40	0.0027	1.5254
23	0.0027	1.1886
124	0.0027	1.7496
16	0.0027	1.2081
65	0.0027	1.0569
96	0.0029	2.264
32	0.0029	1.1692
69	0.003	1.0114
88	0.0031	1.082
29	0.0032	1.161
45	0.0032	1.774
42	0.0033	2.8872
18	0.0033	1.2601
22	0.0036	1.3242
72	0.0036	2.1313
20	0.0036	1.2422
57	0.0036	1.0663
2	0.0037	1.0936
30	0.0038	1.2767
98	0.0039	1.711
37	0.0042	1.1133
6	0.0043	1.3364
85	0.0046	2.2632
25	0.005	1.7122
106	0.0068	1.8086
41	0.007	1.0851
81	0.0076	2.2814
75	0.0078	1.7841
10	0.0094	1.3552
52	0.0097	1.675
53	0.0097	1.0837
68	0.0151	1.3654
13	0.0155	1.3497
1	0.0155	1.6753
19	0.0171	1.9703
34	0.02	1.789
8	0.0433	1.2997
4	0.0538	1.6608
51	0.0592	1.2926
64	0.1278	1.0166

[mcjkula]

There is one "problem" with this criteria though, because NAV in the most simple way means ALPHA in the pool > ALPHA held by users, post-halving each new subnet would naturally lean towards a NAV < 1 given that the ALPHA injected into the pool should be halved, while the ALPHA emitted to participants will stay the same until the ALPHA halving of the subnet.

This makes older subnets more vulnerable to being registered with this criteria.

[pycorn0729]

If the owner already unstaked all of his owner alpha and he get the lock_amount, so the owner can get margianl profit by regging?

[mcjkula]

If the owner already unstaked all of his owner alpha and he get the lock_amount, so the owner can get margianl profit by regging?

For the lock cost refund it doesn't look at what the owner holds, it looks at the emissions of the subnet (lifetime total) and gets only the owner_cut (18%) to determine how much the owner has received. So the owner can't get more because of unstaking, in fact he could potentially get less by unstaking when the liquidation price would be higher in the end.

[Swamination-Yuma]

Would this structure promote bottom-scraping? All the data would be there to pick up pre-dereg subnet tokens for potential TAO payout benefitting from the entire TAO_in history.

[mcjkula]

Would this structure promote bottom-scraping? All the data would be there to pick up pre-dereg subnet tokens for potential TAO payout benefitting from the entire TAO_in history.

If you mean that people could stake into so that they benefit from the higher liquidation price, then yes, I talked about it in CoR before. Esentially "arbitrage"