BIP 77: Delimit fragment params with `-` instead of `+`

nothingmuch commented at 12:48 pm on July 5, 2025: contributor

This PR addresses the concerns raised in #1885 with regards using + as a delimiter, by switching to -. Some URI libraries implement RFC 1866’s (HTML 2.0) “keyword” delimitation in query parameters by decoding + as unconditionally, which necessitates kludgy workarounds.

Additionally fragment parameters must now ordered lexicographically (instead of reverse), the motivation for a canonical ordering was to reduce implementation fingerprinting concerns, but the reverse lexicographical ordering was specified simply because that was the behavior that was already implemented in PDK. Since that is somewhat surprising and the delimiter change breaks backwards compatibility anyway, the marginal cost in terms of compatibility of reducing the surprisingness of the reversed ordering is arguably zero, so this is also addressed here.

cc @kumulynja

in bip-0077.md:742 in 315dcc2252 outdated

737+parameters.
738+
739+As a result of this change implementations are encouraged to still accept `+`
740+delimited fragment parameters for during 2026 (a grace period of just under 6
741+months).
742+

murchandamus commented at 8:04 pm on July 7, 2025:

This part would perhaps better work in a “Changelog” section, given that the Backwards compatibility section usually refers to existing behavior and other proposals.

nothingmuch commented at 11:13 am on July 8, 2025:

Thanks, I forgot about that part of BIP 3 but that does make a lot more sense.

I retroactively called the merged draft 0.1.0, and this updated version 0.2.0, and moved the compatibility grace period recommendation to the section defining the fragment parameters, since that didn’t seem to make sense in the changelog.

nothingmuch commented at 1:13 pm on July 8, 2025:

See also #1891 which only contains what is now the first commit of this PR

nothingmuch force-pushed on Jul 8, 2025

in bip-0077.md:11 in cbf69f333f outdated

 7@@ -8,6 +8,7 @@
 8   Post-History: https://github.com/bitcoin/bips/pull/1483
 9                 https://gnusha.org/pi/bitcoindev/7B11AE34-27A7-46ED-95BF-66CA13BA26F3@ngould.dev/#t
10                 https://gnusha.org/pi/bitcoindev/3C0A6E4C-444E-4E75-829C-1A21D8EE40E0@ngould.dev/#t
11+  Version: 0.2.0

murchandamus commented at 8:17 pm on July 8, 2025:

Per BIP-3: BIP Header Preamble, “the headers must appear in the following order[…]License, License-Code, Discussion, Version, Requires”, so this should be under “License”, but I would suggest to drop the Version header until BIP 3 is activated.

in scripts/buildtable.pl:40 in cbf69f333f outdated

36@@ -37,6 +37,9 @@
37 	'Requires' => undef,
38 	'Superseded-By' => undef,
39 );
40+my %VersionField = (

murchandamus commented at 8:21 pm on July 8, 2025:

The changes to scripts/buildtable.pl were cherry-picked into #1820 as discussed, I think they can be dropped here.

DanGould referenced this in commit 70755a9dac on Jul 9, 2025

Separate fragment params with `-` instead of `+`

Since only Bull Bitcoin Mobile and Cake wallet are currently deployed in
production, both using PDK, and the `+` character is causing some
friction, this change seems justified to avoid similar issues with
future implementations.

43f9688600

nothingmuch force-pushed on Jul 9, 2025

murchandamus commented at 7:07 pm on July 10, 2025: contributor

LGTM, should we wait for a review from @DanGould or is this ready to go?

Nit: We use ISO-8601 formatted dates (YYYY-MM-DD) in the preamble, so I would prefer also using the same format in the Changelog, but either is fine and this is not a blocker.

nothingmuch force-pushed on Jul 10, 2025

nothingmuch commented at 7:20 pm on July 10, 2025: contributor

LGTM, should we wait for a review from @DanGould or is this ready to go?

I’d be more comfortable with an explicit ACK even though the corresponding code changes were merged into PDK

Nit: We use ISO-8601 formatted dates (YYYY-MM-DD) in the preamble, so I would prefer also using the same format in the Changelog, but either is fine and this is not a blocker.

ISO-8601 is better, done

DanGould commented at 7:31 pm on July 10, 2025: contributor

ACK 1474f0d71552b4bc6cc68ec816847be4e487d7ef

in bip-0077.md:744 in 1474f0d715 outdated

737@@ -734,3 +738,15 @@ directory, and development kit may be found here:
738 <https://github.com/payjoin/rust-payjoin>. Source code for an Oblivious
739 HTTP relay implementation may be found here:
740 <https://github.com/payjoin/ohttp-relay>.
741+
742+## Changelog
743+
744+- 0.2.0 2025-06-08

murchandamus commented at 7:56 pm on July 10, 2025:

I was looking at the diff from the force push and noticed that it was July before and now is June. Just wanted to doublecheck whether you meant June.

0-- 0.2.0 Jul 8 2025
1+- 0.2.0 2025-06-08

nothingmuch commented at 8:06 pm on July 10, 2025:

No I’m just an idiot

nothingmuch force-pushed on Jul 10, 2025

Specify lexicographical order for fragment params

The rationale for this change is that since `-` instead of `+` breaks
compatibility anyway, the marginal cost of removing this
unusual/surprising requirement for reverse lexicographical ordering is
zero.

8b83c34949

nothingmuch force-pushed on Jul 10, 2025

murchandamus merged this on Jul 10, 2025

murchandamus closed this on Jul 10, 2025

BIP 77: Delimit fragment params with - instead of + #1890

BIP 77: Delimit fragment params with `-` instead of `+` #1890