ARC0047: Logic Signature Templates

[joe-p]

Provides a standardized way to define logic signatures so wallets can safely inform users what they are signing

[D4Ocrypto]

This great ARC will open the doors to so many new UX improvements! Thank you Joe!

[joe-p]

Before going to final we'll need to add a section for wallet communication (perhaps a new WC method?), but there doesn't seem to be any objections to what is currently in the ARC.

[joe-p]

dApps would need to send wallets both the ARC47 JSON and the variable values. Could be done in a payload like the following that is then processed by the wallet (substitute variables in TEAL, compile, and then sign the program)

{
   "arc47":{
      "name":"Address Opt-In",
      "description":"This program allows a given address to opt in the signer to any asset provided it's approved by the associated application",
      "program":"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",
      "variables":[
         {
            "variable":"TMPL_DELEGATED_OPTIN_APP_ID",
            "name":"Delegated opt-in app ID",
            "type":"application",
            "description":"The ID of the application that will be used for verifying opt ins"
         },
         {
            "variable":"TMPL_AUTHORIZED_ADDRESS",
            "name":"Aurhotized address",
            "type":"address",
            "description":"The address that will be allowed to opt the signer into assets provided it's approved by the associated application"
         }
      ]
   },
   "values":{
      "TMPL_DELEGATED_OPTIN_APP_ID":1337,
      "TMPL_AUTHORIZED_ADDRESS":"46XYR7OTRZXISI2TRSBDWPUVQT4ECBWNI7TFWPPS6EKAPJ7W5OBXSNG66M"
   }
}

[jannotti]

I don't think you need to base64 encode teal programs. They are unicode and so is JSON.

[jannotti]

Why is the type being encoded in hex? It's human readable unicode.

[joe-p]

I think the current wording made it unclear that I was referencing the value rather than the type itself. See if b02fc9c makes more sense

[jannotti]

This example drives home that it's a big ask for the wallet to parse TEAL. According to these rules, it needs to accept:
byte TMPL_NAME; int 7, byteblock "hello\" " TEMPL_NAME for example.

I think we need to spec out the exact local rules that would allow a program to accept. I admit that's not easy.

Perhaps we should restrict things much more strongly. Only things that can be be properly escaped by the wallet, and result in a single byte X or int X instruction, for example

[joe-p]

Yeah didn't think about constant blocks. I think enforcing byte x/int x is reasonable.

I also didn't think about ;. How about enforcing that any usage of a variable must match this regex /^(byte|int) TMPL_/

[joe-p]

Or probably /^\s*(byte|int) TMPL_/ to account for indented lines.

[jannotti]

I'd allow pushint and pushbyte as well. I think there are good reasons to use them in templates, so that the inserted bytes end up exactly where you expect them.

(int and byte do some funny business around optimization, and moving the constants into constant blocks.)

[emg110]

ARC has become excellent, please merge it ASAP Joe

[jannotti]

Do you want to have any rules for the names of template arguments?

[jannotti]

I'd allow pushint and pushbyte as well. I think there are good reasons to use them in templates, so that the inserted bytes end up exactly where you expect them.

(int and byte do some funny business around optimization, and moving the constants into constant blocks.)

[jannotti]

Do we need a .gitignore in here?