> My feeling on this: if a symbol is exported "officially" for use by
> users in their programs, it should work. And if the symbol is not
> intended for use by users, it should not be external in that internal
> package.
Well, that can't be right. SBCL has a right to document packages as
being for internal use, and to export symbols from such packages for its
own internal use. The docstring of SB-INT states (as its first word)
that it is private.
> The use came from PVS, btw, where three packages shadowing-import it
> from sb-int.
The fact that users use it (and perhaps despite it being in a private
package, that it has a documented heritage from CMUCL's extensions)
indicate to me that there is benefit to being robust to potential user
misuse.
"Paul F. Dietz" <email address hidden> writes:
> My feeling on this: if a symbol is exported "officially" for use by
> users in their programs, it should work. And if the symbol is not
> intended for use by users, it should not be external in that internal
> package.
Well, that can't be right. SBCL has a right to document packages as
being for internal use, and to export symbols from such packages for its
own internal use. The docstring of SB-INT states (as its first word)
that it is private.
> The use came from PVS, btw, where three packages shadowing-import it
> from sb-int.
The fact that users use it (and perhaps despite it being in a private
package, that it has a documented heritage from CMUCL's extensions)
indicate to me that there is benefit to being robust to potential user
misuse.