On Mon, Mar 13, 2023 at 2:57 PM Yann Ylavic <ylavic....@gmail.com> wrote: > > On Mon, Mar 13, 2023 at 2:32 PM Eric Covener <cove...@gmail.com> wrote: > > > > On Mon, Mar 13, 2023 at 8:31 AM Yann Ylavic <ylavic....@gmail.com> wrote: > > > > > > Maybe we could make [B=] and [BTCLS] not mutually exclusive (encode > > > both controls and whatever in B=). > > > I was thinking of a [BNEG] flag too (encode everything but what's in > > > B=), and never encode alnum or '_', so all in all some further patch > > > like the attached one. WDYT? > > > > Looks good to me, I have an ancient patch where I did something very > > similar to a copy of int:escape where you could set exceptions in > > subprocess_env. The config is ugly so I never upstreamed it. > > > > if you make the change I can add some tests/doc. Should caution > > against plain [B] and refer to the others in the doc? > > r1902323, thanks for the tests/doc. > IIUC, when the query-string is rewritten, I think we should caution > against using [B] with a redirect (double encoding)
[NE,BCTLS] might be a good option too for redirects that rewrite the query-string, provided the capture comes from r->uri and not r->args already. > and not using > [BCTLS] (or some careful flavor of [B]) for a non-redirect..
