# z/OS DGW Migration FAQ





## Other sources of information

[Migration
redpaper](http://www.redbooks.ibm.com/Redbooks.nsf/RedbookAbstracts/redp4987.html?Open)

[Infocenter](http://www-01.ibm.com/support/knowledgecenter/SSEQTJ_8.5.5/com.ibm.websphere.ihs.doc/ihs/welc6top_mig_ihszos53_ihs_container.html?lang=en)

[Unsupported contributed config migration tool here:
ftp://public.dhe.ibm.com/s390/zos/tools/cmt/](ftp://public.dhe.ibm.com/s390/zos/tools/cmt/)

[Informal document comparing both HTTP servers and their
directives](conversion.html)

[GWAPI / Module development resources](zos_questions.html#MODDEVRES)

[ZOS Questions FAQ](zos_questions.html) for information about IBM HTTP
Server on z/OS.

## URL mapping questions<!-- {#MAPPING} -->

### Problems with double-slashes<!-- {#multislash} -->

Domino by default "simplifies" URL paths that contained double-slashes,
such as http://www.example.com/icons//spacer.jpg. While these do not
affect mapping of URL paths to the filesystem, they can affect other
modules that work on the URI directly, such as the WAS WebServer
Plug-in.

IHS prior to 9.0.5.0 did NOT merge these consecutive slashes, but
9.0.5.0 and later do it by default as well.

Here is a recipe to remove one double-slash from the URL path for 9.0.0.11
and earlier:

Uncomment LoadModule for mod\_rewrite near top of httpd.conf

Once at bottom of httpd.conf and once in each \<VirtualHost\>:
```apache
    RewriteEngine ON
    # Remove first double-slash from the path component of the URL
    RewriteRule ^(.*)//+(.*)$ $1/$2 [PT]
```

## Translation questions<!-- {#XLATE} -->

### Are there any locale (LC\_\*) differences between DGW and IHS?<!-- {#setlocale} -->

IHS doesn't call setlocale(LC\_ALL, ""), so setting LC\_\* variables
doesn't have any direct affect on locale-sensitive standard library
calls.

## CGI questions<!-- {#CGI} -->

### Why isn't \_BPX\_USERID set for my CGI environment?

Domino unnecessarily set \_BPX\_USERID in a CGI's environment even
though it already changed userids. Apache does not currently set this
variable, but you can copy the "REMOTE\_USER" variable to this variable
to allow existing scripts to continue to function unchanged.

```apache
    <Location /cgi-bin/test-cgi>
      # Example of SAF-authenticated resource.
      Authtype basic
      AuthName foo
      AuthBasicProvider saf
      Require valid-user
      SafRunAs  %%CLIENT%%
    
      # Copy REMOTE_USER to _BPX_USERID. Note that this only works when mod_rewrite
      # is configured in "directory" context. You cannot put this outside of
      # Location/Directory, as it will run prior to authentication.
      Options +FollowSymlinks
      RewriteEngine ON
      RewriteCond %{REMOTE_USER} (.+)
      RewriteRule .* - [E=_BPX_USERID:%1]
    </Location>
```

### FSCP and NETCP environment variables

These variables are not set by Apache. If you depend on them being set
as a global configuration for a CGI, set them with `SetEnv`.

### Why doesn't 'PULL PARSE' work in a REXX CGI?<!-- {#CGIPOST} -->

If the POST body is not newline-terminated, PULL PARSE won't read any
data. There is no reason for a POST body to be newline terminated. DGW
had a workaround to accomodate this by adding data to the users POST
body. Apache doesn't modify the body in this way.

### Why do REXX reads of POST data return partial results?

REXX I/O calls that try to read the HTTP POST body in a single read, such as
`read 0 var 13766` are likely to only return partial results or work erratically.
Generally, when reading from a socket or a pipe, I/O routines that read into 
non-trivial sized buffers should always be called in a loop.

Apache communicates with CGI scripts via a pipe, and the writing side of a pipe typically
cannot write more than 4096 bytes before the reading side begins to consume the data.  The POSIX
standard only requires that pipes have a 512 byte buffer.

Below if a brief routine using a 4k chunk size:

```rexx
ReadPostData:
 body = ""                                      /* Initialize buffer    */
 chunkSize = 4096                               /* Input buffer size      */
 Do FOREVER                                     /* Read POST string chunks   */
  chunk = Copies(" ",chunkSize)                 /* Initialize chunk buffer   */
  Address LINKMVS 'IMWXRD chunk'                /* Attempt to read next chunk  */
  Say "BLMWSWRT: Length(chunk)" Length(chunk)
  If RC \= 0 then Leave                         /* All data has been read    */
  body = body || chunk                          /* Collect data         */
 End                                            /* Read POST string chunks   */
 Drop chunk                                     /*               */
Say "BLMWSWRT: RC:" RC "content-length" env.content_length,
 "bytes read:" Length(body)                     /* RC = 8: END-OF-DATA reached */
 Return                                         /*               */
```

### How can a CGI opt out of otherwise configured translation?

Set `CharsetOptions DGWCompat` and then issue a response header of
`Content-Encoding: binary`. This will be filtered out and short-circuit
translation in 8.5.5 and later.

## Authentication, authorization, and access control questions<!-- {#AAA} -->

### What's the equivalent of the UserID directive?

While there is no direct equivalent, if UserID was being used to change
to a different thread identity,
[mod\_authnz\_saf](http://www-01.ibm.com/support/knowledgecenter/SSEQTJ_8.5.5/com.ibm.websphere.ihs.doc/ihs/tihs_safconfigz.html?cp=SSEQTJ_8.5.5&lang=en)
can do similar userid changes.