Re: [Scheme-reports] Libraries that are referenced more than once

To: Marc Nieper-Wißkirchen <marc.nieper@x>

Subject: Re: [Scheme-reports] Libraries that are referenced more than once

From: Alex Shinn <alexshinn@x>

Date: Wed, 29 Jan 2014 11:12:03 +0900

In-reply-to: <CAEYrNrQeFS4HQqtyZuq1jC=RcxRGBNd8UE4vXCF-Jv=jNHGYCQ@mail.gmail.com>

References: <CAEYrNrTk2kmkWjd+nwX000ctjtN3mBdsTF6taV2DJJ3ouUkH0A@mail.gmail.com> <CAMMPzYOpM3tvSFScPNi42Y7KjJr54Eis-xcgQNkEmETDFZ9-NQ@mail.gmail.com> <CAEYrNrTV=qFCLU1FVEz96NtfbdxVnnYYuG_-so_KVnaBZ9icBw@mail.gmail.com> <CAMMPzYMSECz6GKquq8ZOFNuq3nPijTE5aAMFH9uicmUVk5-o8w@mail.gmail.com> <CAEYrNrQzAh8+3rtUphBBDLMWQL7ApjchPCwO0GaLEmy4dhZSNg@mail.gmail.com> <CAMMPzYND50gaAvZSrzMB052ukE1zjy5bsJZdzE-c6vnaynfn8Q@mail.gmail.com> <CAEYrNrSZSwCLxiUAo6Ujq7fXaGw8wjGWTiam7xYi7SZz3wK31g@mail.gmail.com> <CAMMPzYPw3o-MWbpz_K9xzwTTMXm3+oA99Rr_AVeWkoO-NCrnkQ@mail.gmail.com> <CAEYrNrRBhPQUB85k+XEY18_CX1+2wO6tsRTiZ67WK6JAV=rdfg@mail.gmail.com> <CAMMPzYPo7S74WHvbYTMZfE5kf_R_uiB+jfutiPUq0TCVGYEH7A@mail.gmail.com> <CAEYrNrRhTTkrdvi8vtp0Q1=y0JJ4ygXOmoRak3yeMWwHROQzCA@mail.gmail.com> <CAMMPzYOe-NtFgNvAosV0mKkAM-BMUWbiC7WBKw9prNd7KuC2eQ@mail.gmail.com> <CAEYrNrQeFS4HQqtyZuq1jC=RcxRGBNd8UE4vXCF-Jv=jNHGYCQ@mail.gmail.com>

On Tue, Jan 28, 2014 at 11:42 PM, Marc Nieper-Wißkirchen <marc.nieper@x> wrote:

[...]
In this example, `keyword' is bound in `macro-library', so can be exported by `macro-library'. So, in this case, is it true that this program is not portable because the binding of `keyword` in the top-level program may be different than the binding of `keyword' in `wrapper-library' in case `macro-library' is loaded twice?

The report doesn't say much about this, so if you took a liberal

interpretation you might be able to conclude this wasn't portable.

Realistically one would assume that libraries are internally consistent,

and that instantiations of the same library during the same phase

and same compilation would in fact be the same.

Again, we fall into the "difficult to talk about without specifying fully"

territory. You should definitely read what R6RS has to say about

how hygiene works:

http://www.r6rs.org/final/html/r6rs-lib/r6rs-lib-Z-H-13.html#node_sec_12.1

This is not the only hygiene algorithm in use, and the alternatives

may behave differently in corner cases. R7RS will specify explicit

renaming macros as a lowest common denominator low-level macro

library, which hides much of the details and may allow a simple and/or

more flexible specification of hygiene. Syntactic-closures are another

alternative which work differently from the R6RS system, and Andre

van Tonder more recently put a lot of work into another hygiene

algorithm in SRFI 72:

http://srfi.schemers.org/srfi-72/

(In chibi-scheme, the binding apparently is the same because I get the answer 42.

Sure, we are a priori talking about changing the specification to be

more clear on a hypothetical case that no one ever has nor likely

ever would implement.

Chibi specifically fits into the common pattern that no library is ever

loaded more than once.

Alex