Follow-up to 549afce2ee1339370dec52787548645ab1f8c3c5
Add in-repo BIP70 extension registration page #36
pull petertodd wants to merge 1 commits into bitcoin:master from petertodd:bip70-extensions changing 3 files +16 −7-
petertodd commented at 3:41 PM on March 17, 2014: contributor
-
Add in-repo BIP70 extension registration page e2331a1574
-
jgarzik commented at 5:58 PM on March 17, 2014: contributor
Technical ACK -- this is technically true and technically an improvement, but, really this text is talking narrowly about extending the protobufs protocol definition.
More widely, any extensions should be documented in a BIP.
-
petertodd commented at 6:02 PM on March 17, 2014: contributor
Any suggestions on the wording to make that clear?
For now it may be best to just add something along the lines of "This page is only to avoid conflicts; you should also write a BIP document describing your extension as well."
-
gavinandresen commented at 5:19 PM on March 18, 2014: contributor
ACK.
RE: extensions should be documented in a BIP: only if they prove to be useful. I expect there will be "we thought it was a good idea at the time" extensions that never become BIPS.
-
petertodd commented at 7:11 PM on March 18, 2014: contributor
Idea: add an explicit private/experimental-use-only numbering range.
-
jgarzik commented at 7:21 PM on March 18, 2014: contributor
Good point. Most specs that must allocate number/name-spaces include
- An experimental range. Anybody is free to use these without a BIP for experiments etc. If you get clobbered and break something, you get to keep both pieces.
- A vendor-specific range. Any vendor rolling out vendor-specific extensions uses this range.
-
gavinandresen commented at 7:44 PM on March 18, 2014: contributor
Y'all like making things complicated...
RE: experimental use range: NACK on that idea, because then there is pain when a successful experiment "needs" to change numbers to be Officially Supported. See the history of the x-foo "experimental" MIME types for a good example of why not to go that way.
RE: vendor-specific range: I don't like that either, for the same reason (successful vendor extensions get used by multiple vendors, then you have pain if you need to support the same feature as both a vendor and a standard extension).
-
jgarzik commented at 7:47 PM on March 18, 2014: contributor
Well the general point is to let people self-organize, rather than centrally administer.
- gmaxwell referenced this in commit 856bb84058 on Apr 24, 2014
- gmaxwell merged this on Apr 24, 2014
- gmaxwell closed this on Apr 24, 2014
- real-or-random referenced this in commit 844845a6c3 on Aug 10, 2022