- From: Poul-Henning Kamp <phk@phk.freebsd.dk>
- Date: Sun, 12 May 2024 19:27:04 +0000
- To: Roman Danyliw <rdd@cert.org>, Roman Danyliw via Datatracker <noreply@ietf.org>, mnot@mnot.net, phk@varnish-cache.org
- cc: "The IESG" <iesg@ietf.org>, draft-ietf-httpbis-sfbis@ietf.org, httpbis-chairs@ietf.org, ietf-http-wg@w3.org, tpauly@apple.com
-------- Roman Danyliw via Datatracker writes: > -- Section 4.1.5. What is the thinking behind the design to only support a > range of +/- “999,999,999,999,999”? It’s much smaller than an int64. At least one widely used web-programming languages turn everything numeric into IEEE-754 binary64 floating point. That format has a 53 bit wide mantissa, which can safely transport 15 without loss of information. > -- Section 4.1.6. What is the thinking behind the design to only support > “three significant digits to the right of the decimal point” and “12 > significant digits to the left of the decimal point”? Should this be expanded > to at least be equivalent to decimal64 Same as above: 15 decimal significant digits. If people need to move numbers outside these types, there are many ways they can do that, for instance using sf-string or sf-binary serializations. The only examples offered during the WG process, where 15 digits would be insufficient, were all related to cryptography, where sf-binary would be cheaper and safer. -- Poul-Henning Kamp | UNIX since Zilog Zeus 3.20 phk@FreeBSD.ORG | TCP/IP since RFC 956 FreeBSD committer | BSD since 4.3-tahoe Never attribute to malice what can adequately be explained by incompetence.
Received on Sunday, 12 May 2024 19:27:11 UTC