ISO C++ committee has a new chief sheep herder

The ISO C++ Standards committee, WG21, has a new convenor, Guy Davidson, or rather they will have when the term of the current convenor, Herb Sutter, expires at the end of this year.

Apart from the few people directly involved, this appointment does not matter to anybody (sorry Guy). The WG21 juggernaut will continue on its hedonistic way, irrespective of who is currently the chief sheep herder.

Before discussing the evolution of language standards, a brief summary of the unusual points around this appointment:

1. More than one person volunteered for the job (several in the US, who selected Jeff Garland, and one in the UK; everyone agreed that both were capable candidates). The announcement by a programming language convenor that they are not standing again when their 3-year term expires more commonly kicks off discrete discussions about whose arm can be twisted to take on the role. It’s a thankless task that consumes time and money (to attend extra meetings). Also, the convenor has to be neutral, which circumscribes being involved in technical discussion.

   Sometimes an outsider pops up, ruffles a few feathers and then disappears (from the Standards’ world).

2. One of the SC22 (the ISO committee responsible for programming languages) convenor selection rules says (see Resolution 14-04): “When a WG Convenorship becomes vacant, … and multiple NBs have each nominated a candidate, the Convenorship shall be assigned to the candidate whose NB currently has the fewest SC 22 Convenors.” Currently, the US holds multiple convenorships and the UK holds none, so the UK nominee is appointed.

   As often happens, people like diversity rules until they lose out. The US submitted a selection procedural change to SC22, and asked that it take effect before the selection of a new WG21 convenor. The overwhelming consensus at the SC22 plenary last Monday was not to change the rules while an election was in progress. An ad-hoc committee was set up to consider changes to the current rules.

End of the news and back to regular postings.

Standards committees for programming languages are now a vestige from a bygone era. The original purpose of standards was to reduce costs (the UK focused on savings achieved through repeated use of standardized items and the US focused on reduced training costs) by having companies manufacture products that conformed to a single specification.

There were once a multitude of implementations for the commercially important languages, each supporting slightly different dialects (the differences were sometimes not so slight). Language standards provided a base specification for developers interested in portable code to keep within, and that vendors could be pressured to support.

The spread of Open source compilers significantly reduced the need for companies to invest in maintaining their own compiler (there might be strategic reasons for companies selling hardware or operating systems to continue to invest in their own compiler), and reduced the likelihood that customers of commercial compiler companies would continue to pay for updates (effectively driving most compiler companies out of business).

Language standards are redundant in a monoculture, i.e., where only one compiler per language is widely used. For some years now, there have been a handful of actively maintained compilers for the widely used languages.

These days, conformance to a language standard is measured by the ability of an implementation to compile and execute the Open source software available in the various ecosystems.

As has often been observed, committees find work to keep themselves busy, and I have seen announcements for new ISO committees that look like they were created because somebody saw a CV padding opportunity.

I continue to think that the C++ committee has become a playground for bored consultants looking for a creative outlet.

WG21 meeting attendance continues to grow, now attracting 200+ attendees (Grok undercounts, e.g., 140 vs 215, and ChatGPT 5 is completely out of its depth). This is an order of magnitude greater than the C committee, WG14, and in a few years could be two orders of magnitude greater than the other SC22 languages.

The two major C/C++ compiler vendors (i.e., gcc and llvm) could simply go their own way, with regard to new language features. However, I imagine that “supporting the latest version of the language standard” is a great rationale to use when asking for funding.

How large can WG21 become before it collapses under the weight of members and the papers they write?

The POSIX standard, WG15, meetings often had 200-300 attendees in the late 1980s/early 1990s. But the POSIX committee stuck to its goal of specifying existing practice, and so has faded away.

Guy strikes me as an efficient administrator. Which is probably bad news, in the sense that this could enable WG21 to grow a lot larger. What ever happens, it will be interesting to watch.