that's also sounds like a problem one should not run into with a language where the language level primitives are able to do the right thing
yeah, true. Not sure what is going wrong there.
presumably I am doing something stupid somewhere
yeah, finding the root is recommendable
I suspect it is in the link generation for making the static html pages
i keep telling you to not try to roll your own web things :smile:
it took our species some 15 years to get collectively not horrible at that, maybe stand on those shoulders
Ha. Probably just need to set the character set on the generated html index pages
that idea sounds even worse to me
I haven't seen a lot of evidence of "not horrible"
it implies you're doing your own content negotiation scheme somewhere to begin with, and instead of actually handling that for anything a client might request, you just do 'something'
I'll ping you on the code change so you can shake your head at it
i can comment on your content negotiation flow graph if you have one
the markup generation side of it does not even matter if the underlying transport layer is not up to par
The HTML for that page is generated at channel import time, listing the files and directories
and there's no render time transform layer?
and no storage layer semantics around what the payload's pile of bytes is encoded as
nope, the web serving is fairly simple. Look up item for URL path, serve thing
there is size and content type saved
with the thing
yeah, that's http 1.0 era internet stuff
we went quite far from there
does content type include storage encoding details as well?
i guess not really if people "just edit things with text editors and just serve files"
what sort of sort of storage details? There is a map of objects for each channel and if an object exists it maps to a ContnetID, which is a message on a channel block
The message on the channel block might reference other chunks if the file isn't trivially small
details like "is this ascii, latin-1, utf-8 - which byte order - how are composables normalised"
ah, I don't have any of that
could add it as needed
and that's probably why you don't have an oracle based on which to set the correct response header for the content you serve
so am setting the content-type based on metadata. I can add other metadata as needed.
nor the ability for a client to request it in an another one and that being transformed on the fly
content type charset is what i'm mostly on about
hardcoding that is an equally bad idea to not setting it at all
yeah, I can see that
and going deeper into getting that right is something i do not recommend rolling your own on
and when i say web, you think it's only http, whereas i take the tim berners lee holistic view of the total service
the combination of transport and content and client empowernment is a classic and difficult triangle to fully cover
when you do "web", you're not "serving files" or generating "api responses" anymore
content is hard
and we're not even talking of semantic web here yet :slightly_smiling_face:
Right now, I'm focusing on simple have static content, want to serve static content sort of cases
if you just want to do serving of files, maybe find a WebDAV implementation
Serving files to browsers mostly :wink:
I remember when webdav was hot
it never went anywhere
that even half looks like what you're cooking