EDNS0 interface needs to be exported

[miekg]

The EDNS0 interface is

type EDNS0 interface {
	// Option returns the option code for the option.
	Option() uint16
	// pack returns the bytes of the option data.
	pack() ([]byte, error)
	// unpack sets the data as found in the buffer. Is also sets
	// the length of the slice as the length of the option data.
	unpack([]byte) error
	// String returns the string representation of the option.
	String() string
}

This means I can't externally define a new option; i.e. a bogus 100 one to do some EDNS0 compliance testing.

[tmthrgd]

This seems sensible to me. Should pack take a []byte to allow it to avoid allocation and be more like PrivateRR?

[miekg]

[ Quoting <notifications@github.com> in "Re: [miekg/dns] EDNS0 interface nee..." ]

This seems sensible to me. Should `pack` take a `[]byte` to allow it to avoid allocation and be more like `PrivateRR`?

Maybe, don't you then need to know the size before hand?

[tmthrgd]

It would need a Len method then. Right now (*OPT).len calls pack just for the length.

PrivateRdata has one:

// PrivateRdata is an interface used for implementing "Private Use" RR types, see
// RFC 6895. This allows one to experiment with new RR types, without requesting an
// official type code. Also see dns.PrivateHandle and dns.PrivateHandleRemove.
type PrivateRdata interface {
	// String returns the text presentaton of the Rdata of the Private RR.
	String() string
	// Parse parses the Rdata of the private RR.
	Parse([]string) error
	// Pack is used when packing a private RR into a buffer.
	Pack([]byte) (int, error)
	// Unpack is used when unpacking a private RR from a buffer.
	// TODO(miek): diff. signature than Pack, see edns0.go for instance.
	Unpack([]byte) (int, error)
	// Copy copies the Rdata.
	Copy(PrivateRdata) error
	// Len returns the length in octets of the Rdata.
	Len() int
}

[miekg]

[ Quoting <notifications@github.com> in "Re: [miekg/dns] EDNS0 interface nee..." ]

It would need a `Len` method then. Right now `(*OPT).len` calls `pack` just for the length.

Ack. That seems to make sense then.

[tmthrgd]

#1041 was reverted, see #1045 and #1046. Reopening for decision.

[micahhausler]

I'm developing a CoreDNS plugin and have a need to set a non-spec'd EDSN0 option, I see that the interface was made exportable and then reverted, is there a plan to make this breaking change tied to a future release? Is there a workaround for this besides forking this repository?

[miekg]

You probably don't need that to just have a custom EDNS0 option.

…

On Tue, 24 Mar 2020, 18:31 Micah Hausler, ***@***.***> wrote: I'm developing a CoreDNS plugin and have a need to set a non-spec'd EDSN0 option, I see that the interface was made exportable and then reverted, is there a plan to make this breaking change tied to a future release? Is there a workaround for this besides forking this repository? — You are receiving this because you modified the open/close state. Reply to this email directly, view it on GitHub <#857 (comment)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/AACWIW5U2WYSJSNUY7X6HYTRJDVADANCNFSM4GHUAWOA> .

[miekg]

We've tried, maybe too hard, maybe something for a v2....