Hi, On Wed, 2026-07-29 at 02:15 +0100, Steve McIntyre wrote: > On Tue, Jul 28, 2026 at 09:39:57PM +0200, Marc Haber wrote: > > I do not think that is fine, though. Why would I put a sound suggestion on > > the bottom of the pile or bin it just because it was written with the wrong > > editor or using a mail service I don't like? > > This is a false equivalence, again. > > Sending mail via gmail or using vim to write your code don't > materially affect the quality of that code. Neither are likely to > cause hallucinations in the code, or to use vast amounts of resources > in its production. (etc.) > > There is a substantial difference here, and that's why people care > about this so much.
Gmail would very specifically be one of the tools that some people want to ban for contributions... Specifically any use of LLMs for assistance with communication was a point of contention. So Gmail-originated mails would be banned or need to be marked and ignored by some maintainers, depending on the resolution, even when only being used for stylistic improvements. The bans also apply without any regard for quality or presence of hallucinations. We specifically also ban LLM uses that materially improve the quality of code or remove hallucinations present in the code. Why would a LLM removing hallucinations be bad? Note that LLMs can improve code quality is a simple fact. Denying this and claming LLM use would always lead to hallucinations or worse code quality is just wrong, i.e. a hallucination. Resource usage is also a personal choice currently. We do various things that could be avoided to reduce that (DebConf travel, number of architectures, QA rebuilds, reproducible builds, long support time for old releases, number of install images to build, ...), yet we keep doing all of them instead of having individuals forcing their preferred resource usage constraints on other people. How long would one have to use LLMs to consume an amount of resources that is comparable to the vast amount consumed by a DebConf attendance? If people who make this argument can't answer that, I would like to ask them to apply consistent standards. Note that the total amount of resources to construct something also isn't really relevant: the resource cost for the specific use plus a proportional amount of the construction/maintenance costs is a much more reasonable measure. Otherwise for a intercontinental flight, one would end up including the total resources required to build the plane, airports and supporting infrastructure (airports required for emergency landings, radar, ...). Ansgar
