Comment 1:

Perhaps a bool on the Encoder struct.

Labels changed: added repo-main.

Comment 2 by wangz@google.com:

Automatic HTML-escaping is irrelevant to JSON encoding. Therefore the feature should be
turned off by default.
The "auto-escaping is safer" argument in the original thread
(https://golang.org/issue/3127) is bogus because it all depends on
the context where the serialized data is used.
Before this gets fixed, one quick workaround is to fork the encoding/json module and
remove "&& b != '<' && b != '>' && b != '&' " from the following two functions in
encode.go:
func (e *encodeState) string(s string) (int, error)
func (e *encodeState) stringBytes(s []byte) (int, error)

Comment 3:

The result is valid JSON. The details of how it gets encoded are not guaranteed and
don't need to be under user control. The only guarantee is that the result is valid and
semantically correct JSON, which it is. It is not worth making the API more complex to
control whether < > get escaped. If you really care, do a global search and
replace on the output to put them back.

Status changed to WorkingAsIntended.

Comment 4 by wangz@google.com:

The result is valid json but also unnecessarily inflated in many contexts where '<',
'>', and '&' abound.
Doing string replacement in a second pass is an inefficient and cumbersome workaround.
Please consider Brad's suggestion to add a bool to the Encoder struct.

Comment 5 by karim.nassar@vervemobile.com:

+1 for flag. In my context, I am generating the html that is in response to an api that,
in some cases, mismanages the escaped text. I can't control them, and I trust myself.
Forking the json pkg seems pretty heavy weight, and string replacement is woefully
inefficient when milliseconds matter.