Directory Conventions: Complete Version

From DLXS Documentation

Contents 1 Rationale 2 Conventions for Content and Middleware 3 Conventions for Software Packages 3.1 Basic Convention 3.2 Special Cases 4 Data Flow Describing the Preparation and Release of a Collection

[edit] Rationale

Multiply the amount of data you have by the number of people dealing with it and you get a metric of complexity indicating how badly you need conventions. Having some basic directory conventions solves some general problems:

• things are easier to find (e.g., for DLXS distributions)
• things are easier to move
• it's easier to tell when you have found the thing you're looking for (i.e., if it's in the right place, it generally must be the authoritative version)

With respect to content and middleware, a good convention:

• protects data (e.g. XML, images), indexes, and CGI middleware from being accidentally served to a client by the web server
• protects static HTML and XML from CGI-related security holes
• insulates the web server software from the content and middleware, which makes web server upgrades easier

From a system administration point of view, a good convention:

• provides a structure that's pretty easy to traverse and can scale up without getting leggy
• allows for the coexistence of different classes and collections on a single server

[edit] Conventions for Content and Middleware

Content should be organized by DLPS identifier or, if that's unavailable, collection identifier. Middleware and everything else should be organized by collection identifier.

The basic scheme is as follows:

<prefix>/bin/c/class or <prefix>/bin/c/collection	loading/munging/processing scripts optionally by class or collection, if appropriate typically on the staging server, not production servers
<prefix>/cgi/c/class or <prefix>/cgi/c/collection	CGI middleware by class or collection
<prefix>/img/c/collection	continuous tone image data non-web-ready (archival) images, to be delivered through middleware; i.e., like image services web-ready (GIF, JPEG, PNG) images which are licensed content but to be delivered directly; to make this work, there is typically a symbolic link from <prefix>/web; see below
<prefix>/idx/c/collection	search indexes and index-related files
<prefix>/lib/[c/class] or <prefix>/lib/[c/collection]	CGI middleware library modules optionally by class or collection, if appropriate
<prefix>/misc/[c/class/] or <prefix>/misc/[c/collection/]	additional files for DLXS distribution that don't belong elsewhere: e.g., DTDs, character entity sets, catalogs, etc. optionally by class or collection, if appropriate
<prefix>/misc/db	text-based databases: e.g., DLXS sessions database in CSV format
<prefix>/misc/[c/class/]maps or <prefix>/misc/[c/collection/]maps	region map files optionally by class or collection, if appropriate
<prefix>/obj/c/collection	production content organized by collection identifier .raw for "raw" (i.e., unprocessed OCR) text .sgm for "cooked" (i.e., proofed and encoded) text .tif for page images, etc.
<prefix>/obj/d/l/p/dlpsid.vvvv.iii	production content organized by DLPS identifier same file naming conventions as above
<prefix>/prep/c/collection	pre-production (munge-in-process) content organized by collection identifier same file naming conventions as above typically on the staging server, not production servers
<prefix>/prep/d/l/p/dlpsid.vvvv.iii	pre-production (munge-in-process) content organized by DLPS identifier same file naming conventions as above typically on the staging server, not production servers
<prefix>/web/c/collection	static HTML pages
<prefix>/web/c/class/graphics or <prefix>/web/c/collection/graphics	user interface glitter: banners, icons, logos, etc. optionally by class or collection, if appropriate
<prefix>/web/c/class/images or <prefix>/web/c/collection/images	web-ready image content served directly, usually figures that appear in text collections if image content is licensed, files are stored in <prefix>/img (see above) and this is just a symbolic link optionally by class or collection, if appropriate

<prefix> varies from installation to installation. All DLPS servers use at least the prefix /l1; most have additional file systems named /l2, /l3, etc.

Note that although classes and parts of the organization can also exist at this level (for example, h/hti for the Humanities Text Initiative or i/is for Image Services), the individual collections would exist within their own spaces, i.e., <prefix>/.../m/musart.

Personal directories can also exist at this level, i.e., <prefix>/cgi/c/csnavely, for areas to do testing in.

The organization of files below the levels described here is outside the scope of this convention. We don't feel a need to prescribe a structure for inside /l1/web/b/bas, for example. Of course, consistency between like things is good practice.

[edit] Conventions for Software Packages

[edit] Basic Convention

The DLPS institutional convention is that software should be installed as

/l/local/package-version

where

package is the proper name of the software package, preferably lowercase (e.g. "apache"), and version is the full version number (e.g. "1.3.9"), if available.

Likewise, source code can be found in

/l/local/src/package-version

or compressed up in a tarball named

/l/local/src/package-version.tar.gz

[edit] Special Cases

For cases where a version-independent path is needed or wanted, a symbolic link should be created:

/l/local/package -> /l/local/package-version

For commonly-used utilities that are conventionally installed in places like /usr/local/bin, a symbolic link should be created in /l/local/bin:

/l/local/bin/command -> /l/local/package/[subpath/]command

if present, or

/l/local/bin/command -> /l/local/package-version/[subpath/]command

[edit] Data Flow Describing the Preparation and Release of a Collection

Looking at the conventions from a "process" point of view, the following diagram illustrates how we use the directory conventions in the workflow of preparing and releasing a collection.