On Tue, Oct 13, 2026 at 02:25:02PM +0300, Eli Zaretskii wrote: > > Date: Mon, 12 Oct 2026 17:56:45 +0200 > > From: Patrice Dumas <[email protected]> > > Cc: [email protected] > > > > > > Scanning the whole file, however, does not have any visible effect, save > > > > time passed and electricity consumption. > > > > > > Scanning a 5MB file each time the user types "i <something> RET" is > > > not fun. I'd prefer having "Index" untranslated any day. > > > > In the standalone Info reader, it is not each time, concatenated indices > > are cached until the file for which the index is searched changes. > > Cached for a single manual, right? What if I change to another > manual, then back, etc.? This happens a lot for me, for example, when > reading the Emacs manuals, which cross-reference each other. > > And even the first time, it's not fun.
Info should be a fast program even on old hardware. If you think there is unnecessary processing going on when loading a new manual which is causing a noticeable delay, please let us know. I understand your complaint about a lack of "fun" is hypothetical and not your current Info experience. As far as I understand, we are in agreement not to scan the whole Info file for the special markers when loading a new manual. > > But I get your point. > > > > Notice that having Index untrlanslated will always be possible in my > > proposal anyway and be tried first, such that the search can be > > accelerated by putting untranslated "Index" in the node name. > > I'm saying that moving to translated Index nodes needs to avoid > slowing down the reader, in order to take traction. >
