Repository navigation
TIP-795 Discussion: Username Registration Rules #797
Copy link
Copy link
Closed
Description
TIP-795 Discussion: Username Registration Rules
Summary
This issue opens community discussion for TIP-795 (Username Registration Rules). The proposal introduces a native username system for TRON accounts with two categories:
- Short usernames (≤ 8 characters): premium, transferable, subject to an annual maintenance fee; initial allocation managed by TRON DAO.
- Regular usernames (≥ 9 characters): free to register, non-transferable, no maintenance fee.
Motivation
Human-readable usernames reduce user mistakes (copy/paste errors) and improve UX and adoption. This issue invites community feedback to refine rules, security, governance, and implementation details.
Specification (short)
- Allowed characters: ASCII letters (A–Z, case-insensitive) and digits (0–9). No Unicode or special characters.
- Length rules:
- ≤ 8 chars: premium, initial sale/auction by TRON DAO, annual maintenance fee, transferable and tradable on secondary markets.
- ≥ 9 chars: free, register-first-come-first-served, non-transferable.
- Name change policy: one name change per account per year.
- Governance: TRON DAO manages pricing/auctions and maintenance fees for premium names.
- Backward compatibility: usernames are an alias layer; existing base58 addresses remain valid.
Discussion points / Questions for the community
- Are the character and normalization rules sufficient to prevent phishing (e.g., homograph attacks)? Suggestions: forbid similar-looking characters, normalization rules, or a blacklist?
- Should premium name allocation use auctions, lotteries, or other mechanisms?
- What is an appropriate model for the annual maintenance fee? (fixed fee, dynamic, DAO-voted)
- Transfer and marketplace rules — any constraints or recommended on-chain mechanisms?
- Implementation: protocol-level vs smart-contract – pros/cons for TRON ecosystem?
- Any regional, legal, or trademark considerations we should add (brand protection, dispute resolution)?
Implementation notes
- A TRC (or protocol-level) mapping
username → account addressis proposed. Normalization to lowercase for matching. - New RPC endpoints for resolution (username → address) would be needed; wallets/exchanges adoption required.
Next steps
- Please comment with feedback, objections, or suggested edits below.
- After community consensus and addressing major concerns, TIP-795 PR (TIP-795: Username Registration Rules #795) will be updated / approved and proceed toward merge.
Links
- PR: TIP-795: Username Registration Rules #795
- Previous draft: TIP-794 (superseded)
Author / Contact
Thanks — looking forward to community feedback!
Reactions are currently unavailable
Metadata
Metadata
Assignees
Labels
No labels
Activity
alichatme commentedon Sep 29, 2025
Hello TRON community 👋
This Issue is opened for discussion of TIP-795: Username Registration Rules (#797).
Summary of the Proposal:
Motivation:
The main goal is to improve user safety by reducing copy/paste mistakes in addresses, and to make TRON accounts easier to use and remember.
We welcome feedback, suggestions, and discussions from the community.
Your input is very valuable to make this proposal stronger. 🙏
Thanks for your time!
barbatos2011 commentedon Sep 30, 2025
It may be worth considering whether the existing on-chain account_name can be reused, with its features extended or propagated.
For the “short name” design, implementing it as a smart contract may be preferable, as it would provide richer functionality and allow more frequent updates.
Further discussion is needed to determine whether a unified approach could cover all these cases, which would be the ideal solution.
In addition, more background information is needed about the scenarios in which the username can be used — specifically, when, where, and how the name should be applied. This will help the community design it in a more suitable and effective way. Very glad to see more discussion happening here!
halibobo1205 commentedon Sep 30, 2025
Apart from the above
account_name, a similar fieldaccount_idis also provided. Please review it first to see if it meets the requirements.alichatme commentedon Sep 30, 2025
alichatme commentedon Sep 30, 2025
L-Reid commentedon Oct 10, 2025
It would be preferable to make use of the existing fields where possible. Otherwise, introducing new fields that involve protocol changes could be quite risky.
alichatme commentedon Oct 10, 2025
alichatme commentedon Nov 13, 2025
Thank you everyone for your valuable feedback and participation in this discussion.
Based on the extensive community feedback and structural improvements, the proposal has been updated and migrated to a new issue for continued discussion and review.
👉 Updated Proposal: TIP-796 (New Issue)
👉 Related Pull Request: #803
This issue will now be archived and marked as superseded by the updated version.
The new version reflects major refinements:
TRuppercase prefixWe warmly invite everyone to continue the discussion in the updated thread.
Your insights have been incredibly helpful and deeply appreciated 🙏
Thanks again to all contributors who supported the evolution of this proposal.
Redchar1992 commentedon Jun 5, 2026
Since this discussion has been superseded by #799, I suggest marking this issue as superseded/closing it to keep future feedback in one canonical place.
The remaining username-standard questions — transferability, anti-squatting, wallet display rules, and registry implementation — are easier to track if they all live under the updated TIP-796 thread.
waynercheung commentedon Sep 3, 2026
The author previously confirmed that this proposal was revised and migrated to #799. PR #795 has since been closed, and the current revised proposal is tracked in #799 and PR #920 - the
tip-796.mdfrontmatter in #920 also listsdiscussions-to: https://github.com/tronprotocol/tips/issues/799, so #799 is the canonical discussion thread.@alichatme could you close this as superseded, or could a maintainer close it as a duplicate of #799? The thread remains readable after closing, so the historical context is preserved while further feedback stays centralized in one place.