Comment by zzo38computer

I think you might have misunderstood what I meant, because I was unclear. I meant that it would have to decode the entire EDN string literal containing the base64 data before decoding the base64, not that it would have to decode the entire file before doing so. (I might still be wrong.)

Specifically, I refer to what is quoted below:

> Upon encountering a tag, the reader will first read the next element (which may itself be or comprise other tagged elements), then pass the result to the corresponding handler for further interpretation, and the result of the handler will be the data value yielded by the tag + tagged element, i.e. reading a tag and tagged element yields one value.

> If a reader encounters a tag for which no handler is registered, the implementation can either report an error, call a designated 'unknown element' handler, or create a well-known generic representation that contains both the tag and the tagged element, as it sees fit. Note that the non-error strategies allow for readers which are capable of reading any and all edn, in spite of being unaware of the details of any extensions present.

Due to these things, EDN does not have a proper "octet string" type, even if the extension is added.

> This implementation does not do streaming for now, but it understands a concept of "reading one complete" element from buffer. The only missing part is buffer managment.

OK, then it could be improved.

> Could you explain how it is restricted if you are allowed to run whatever you want during reading of edn document? You can even do IO, no restrictions at all!

Perhaps the above explains how it is restricted. It does not prevent you from looking up data in a database, etc; it is the data model of EDN itself which is restricted; it is not restricting what you do with it.

> Please show how it does better. I'm very curious

Since all types using the same framing, you can do "lazy decoding" if appropriate (you can also use custom decoders in any part of the file, and this can depend on the schema), and ASN.1 does have a built-in octet string type (as well as bit string, unrestricted character string, etc), and you can add implicit or explicit tagging (I prefer to use implicit if the underlying type is sequence or octet string, and explicit otherwise), as well as types such as External (and the nonstandard ASN1_IDENTIFIED_DATA type), you can easily define any type and can easily skip past any field of any type.

> #init/postgres {:db-spec {:host "..." :port 54321 ,,,} :specs {:user ,,,}} [#user/id 1 #user/id 2 #user/id 3]

Even with TER (the below does not use any extensions to TER itself, but extensions to TER are also possible; even if not, whoever reads the resulting DER can handle the application-specific types as needed), you can:

  [ [P:(database.example) 54321] [0A:1 0A:2 0A:3] ]

In this case, the "0A:" prefix means application type 0, which has a meaning specific to the application; presumably for this application, application type 0 would correspond to user IDs. This example uses implicit types for the user IDs; if you want explicit types instead, then you can write:

  [ [P:(database.example) 54321] [0A[1] 0A[2] 0A[3]] ]

Or, if you want to extend TER instead, then you might define your own keyword, e.g. "userid{1}" instead of "0A:1" or "0A[1]".

(TER is not one of the official ASN.1 formats; it is one that I invented for the purpose of having a text format for ASN.1 which can then be converted to DER; most programs would be expected to use DER rather than TER.)