Add definition for plain text

[staab]

No description provided.

[vitorpamplona]

Looks good, but we really need to stop calling this "plain text". It doesn't make any sense.

Adds:

• words starting with #([^\s!@#$%^&*()=+./,\[{\]};:'"?><]+) represent pointers to the t-tagged hashtag of it's name
• words starting with lnbc represent lightning invoices
• words starting with lnurl1 represent lightning wallet uris
• words starting with lno1 represent lightning offers
• words starting with cashuA and cashuB represent cashu tokens.
• words starting with $<a-zA-Z0-9> represent a stock ticker or an asset price.
• words starting with !<geohash> represent a pointer to a g-tagged geohash feed.

Additional rules:

• Urls with scheme, like https://, are expected to be previewed in cards, images, audio players, and video playbacks with the link hidden from the user.
• domain names without schemes, like nytimes.com, are expected to render as is, without previews.
• nip19 uris are expected to either be previewed by the app as a quote OR be a link to an outside app that can view them using NIP-89/typed-scheme (Typed URI schemes. #1539)

[staab]

Looks good, but we really need to stop calling this "plain text".

I thought about introducing the term "nostr flavored markdown" but I think that communicates the wrong thing. "Human writable text" is the closest name I can come up with to the actual justification for this, but it also makes no sense.

Those are good additions, I've added most of them (although making the language somewhat less prescriptive). I left out geohashes since those aren't really human readable or writable. Topics were already in the list.

[vitorpamplona]

"Human writable text" should not include any NIP-19 URIs IMO. Either way, it should be either all or nothing. Keeping very long unreadable, unwritable URIs like nprofile1, nevent, naddr, lnbcs, cashuB and not accepting others like nembed doesn't make much sense.

[staab]

Humans copy/paste urls, they can copy/paste nevent1s too (and they do). They're all the same thing — a URL, i.e. a reference. base64 encoded images and nembeds are content, and while yes, you could copy/paste those too, it's not idiomatic to (for example) copy a pdf's contents into an email instead of attaching it.

[fiatjaf]

I've never seen this consistently implemented anywhere, and I don't think it's wise to specify and conjure it into existence

[staab]

It's less about it being implemented (since there's not really anything a parser really has to do), than being commonly used in rich text editors, and something users do anyhow.

[fiatjaf]

do they? I think users do lists in many different ways, like 1), 1., 1:, or *, -, . they also do multi-level lists.

[staab]

That's a good point. clients should also parse those things probably. I've changed the language slightly so that the given list is what clients should add on behalf of users, but clients will have to parse a much wider range of nonsense.

[vitorpamplona]

Can we add that lists only hold one paragraph?

Meaning that the parser doesn't need to connect lists that are more than a new line away from each other.

This would not work, for instance.

• Item 1: ... ... ...
  .. asdjfasdf
  asdfasdf
  asdfasdf
  asdf

• Item 2

• Item 3

[fiatjaf]

Although I mostly agree with @staab's rationale about human-writable and copyable text, and after lefting a bunch of comments, my conclusion is that actually we shouldn't specify any of these, and the only thing we really should do is follow NIP-27.

Maybe mention somewhere how URLs and image URLs are usually handled, as the behavior regarding these is fairly consistent and expected.

Nostrudel adds a special link every time a NIP is mentioned, should that also be standardized? I don't think so. The same principle applies to $, *, >.

Emojis are also specified somewhere else and very optional. I live very well (better) without them.

[vitorpamplona]

People are copy-pasting nembeds between a DM chat with their doctor to a DM with a pharmacist or an optician.

But I digressed.

@fiatjaf I somewhat disagree. The lack of definition of what is expected of clients to render each kind is a massive UX problem. If some clients show a custom emoji and others don't, that is a problem. The scope kind, say kind 1, must decide to either use or not use custom emojis. But once the decision is made, all kind 1 clients, must follow. There is no option to not implement it. Same for rendering ln invoices, tokens, nip19 uris, etc.

The more we spec what is expected to happen in each "scoping kind" (which defines the use case), the better it gets for users.

[staab]

If some clients show a custom emoji and others don't, that is a problem.

I think it's fine. Coracle doesn't render emojis, and while it probably should, it really doesn't bother me as a user. I understand what they mean by the shortcode. Same for lnurls, cashu, urls, nip19, whatever. In fact, I use coracle in a no-media mode. Clients should have total leeway to render notes however they (their users) want.

[vitorpamplona]

If that is fine, then everything is fine. We can do a full markdown post on Kind 1 and your users will still be able to read it. Which has been what Amethyst has done for years now. If there are no complaints, we can just put markdown (or any other readable format) on every kind.

[vitorpamplona]

For the name, how about "Nostr Structured Text" or "Structured Note Format"?

[staab]

Let's worry about the semantics for now and change the name later. "plain text" is used all over the place already, I'm just trying to define it.

[fiatjaf]

If that is fine, then everything is fine. We can do a full markdown post on Kind 1 and your users will still be able to read it. Which has been what Amethyst has done for years now. If there are no complaints, we can just put markdown (or any other readable format) on every kind.

This is not true. While things like Markdown code blocks don't hurt, Markdown links are horrible for reading anywhere that doesn't support Markdown natively, same for Markdown tables, and even those Markdown titles with leading #. I don't believe you don't understand this.

What about HTML? Or LaTex, or the PDF syntax, I don't know. Any client can decide to support anything, and we certainly shouldn't expect all clients to render all these formats, but also users probably cannot read any of them comfortably.

But users can read an :alpaca: emoji blob and imagine an alpaca just fine.

[fiatjaf]

By the way, the initial Markdown proposal doesn't mention anywhere the horrible []() links or the # titles.

https://daringfireball.net/projects/markdown/index.text

See how the links are much more human-readable with the actual URLs going after the text in their own lines, and the titles are much better too. That's why Markdown was said to be "plain text with an optional rendering step", but of course soon afterwards it got corrupted, we shouldn't be complicit in that corruption.

[fiatjaf]

In any case I'm ok with merging this as an experiment.

[vitorpamplona]

I agree about the[]() links. But you realize how arbitrary that point is right? That's my problem with leaving things open and why being highly prescriptive is awesome.

If we don't provide any guidance, every dev will choose their own arbitrary line. If we do provide guidance (like what this PR is doing), then devs have at least something to be based on...

[cypherhoodlum]

What about markdown tables such as:

Column 1	Column 2
Cell 1, Row 1	Cell 2, Row 1
Cell 1, Row 2	Cell 1, Row 2

Are these human readable enough without being displayed as an actual table? Bigger the table the harder it becomes. Worth a mention in the spec?

[staab]

I don't think tables are easily readable. They're also not "light formatting", they're really structured data.

[vitorpamplona]

We have tables at home.

The tables at home:

Column 1       | Column 2
---------------|----------------
Cell 1, Row 1  | Cell 2, Row 1
Cell 1, Row 2  | Cell 1, Row 2

[fiatjaf]

first column	second column of what is considered a valid markdown table	third column	final row with sums
as you can see	second value of the first row	third value of the first row	12312
this is not very readable	more values here	another value	673
another	row	messy	2316
this is not really plaintext by any stretch of imagination, it's not readable unless it's formatted, so this defeats the purpose entirely

|first column|second column of what is considered a valid markdown table|third column|final row with sums||
|-|-|-|-|-|
|as you can see|second value of the first row|third value of the first row|12312||
|this is not very readable|more values here|another value|673||
|another|row|super|messy|2316||
|this is not really plaintext by any stretch of imagination, it's not readable unless it's formatted, so this defeats the purpose entirely|||||

[mikedilger]

NACK

I think this idea is too proscriptive.  PLANTEXT can be > Anything, > Nothing, > Made Up Formatting with
__NO__ particular predefined meaning.

I know you want to parse and interpret plaintext to render it "better" but I think that is more dangerous than not (rerendering plaintext may totally screw up ascii art, for example).

[staab]

I know you want to parse and interpret plaintext to render it "better" but I think that is more dangerous than not (rerendering plaintext may totally screw up ascii art, for example).

That's what backticks are for. Anyway, this is unavoidable, clients already parse note content and render it to a greater or lesser extent. This PR is about encouraging people to opt for "lesser"