FramerD	is a free-form distributed object database which forms the basis of beingmeta's semantically enabled services and applications.
	[About Us][Technology][Credo][Team][News][Publications][Contact]

---

NEW!beingmeta recently released version 2.6 of the Framer_D database, including numerous refinements and improvements, under both the GNU GPL and LGPL open-source licenses.
[announcement] [download] [release notes] [homepage]

FramerD is a free-form distributed object database supporting millions of objects and optimized for pointer-intensive structures and operations. FramerD includes a variety of inference procedures and a scripting language based on R4RS scheme with special extensions for *non-determinstic programming* and sophisticated text and document handling. FramerD comes with a high performance Apache-based servlet architecture which allows responsive web-based applications and template based scripting through interpreted FDXML files.

Conventional databases based on relational algebra (any SQL-oriented product) are not optimized for pointer-intensive data and poorly support knowledge-based algorithms which make heavy use of pointer-following and queries with complex internal dependencies. FramerD uses an alternate model supporting billions of unique objects with lightweight pointers and a range of caching strategies for resolving object pointers to complex structured objects. This model is incorporated with the ability to manage very large dynamically modifiable indices that map arbitrary structured objects to collections of object references.

Framer_D was originally implemented by Dr. Haase at the MIT Media Laboratory over a decade ago, and has constituted a stable and evolving platform for a range of applications at MIT and elsewhere. beingmeta has continued the development the open-source FramerD platform, focusing on robustness, performance improvements, platform portability, and the FDXML/fdservlet architecture for knowledge-based web applications. The company has also focused on making FramerD able to take advantage of modern SMP hardware. Internally, the company is actively developing entirely new versions of Framer_D for both mobile embedded platforms and high performance hardware.

Framer_D has an associated scripting language, called FDScript, that started as a simple way of specifying inference algorithms but eventually grew into a complete implementation of the Scheme programming language with special provisions for writing non-deterministic programs. These provisions incorporate search processes into the programming language itself, allowing a procedure or database access to return multiple alternative values and to have the language automatically explore the various alternatives independently. For language aficionados, this can be thought of bringing "Prolog-style programming" to a traditional applicative language. In addition, FDScript was extended with rich tools for pattern analyzing textual documents and for connecting to a range of Internet services.

Framer_D 2.6 can be downloaded from Sourceforge ( www.sourceforge.net)

Kind:	File:	URL:
[Source]	framerd-2.6.1.tar.gz	http://prdownloads.sourceforge.net/framerd/framerd-2.6.1.tar.gz?download
[WIN32]	fdinstall261.exe	http://prdownloads.sourceforge.net/framerd/fdinstall261.exe?download
[OS X]	FramerD.dmg	http://prdownloads.sourceforge.net/framerd/FramerD.dmg?download
[x386 RPM]	framerd-2.6.1-1 .i386.rpm	http://prdownloads.sourceforge.net/framerd/framerd-2.6.1-1.i386.rpm?download
[Source RPM]	framerd-2.6.1-1.src.rpm	http://prdownloads.sourceforge.net/framerd/framerd-2.6.1-1.i386.rpm?download
[x386 DEB]	[framerd_2.6.1-3 _i386.deb	http://prdownloads.sourceforge.net/framerd/framerd-2.6.1-1.i386.rpm?download

Framer_D 2.6 includes various bug fixes and expanded error handling but significantly adds:

• New kinds of file indices which scale effectively to multi-million key tables
• General improvements to FDSCRIPT module system
• Defined more core primitive inference procedures especially around structured (assoc/pair) slots
• Added generic facility for file/module reloading
• Various frame operation optimizations
• Added new primitives for accessing serial ports (including USB serial ports)
• Numerous fixes and additions to arithmetic functions in FDScript
• Improvements to index caching performance
• Added compound indices which combine multiple indices with a shared cache.
• Added subservers on index servers, allowing a single server to provide several distinct index tables
• Provided C API access to ints and strings in config files
• New, simpler command line parsing API
• Use long longs as OIDs if possible
• Added support for largefiles (> 2GB) for pools and indices
• Added synchronization for pools and indices
• Use file memory mapping for file pools and indices for dramatically improved performance in many cases
• Further optimiziations to choice/set handling
• Added "schema slotmaps" to reduce memory footprint for homogenous databases
• Extensions to support extensive malloc debugging
• Introduced interned ZSTRINGS as a core data type
• Extended the frame data model by adding *adjuncts* which allow index mappings to specify particular slots of for particular pools.
• Separated out get-methods and compute-methods
• Different pools and indices on the same server now get their own network connection to reduce thread contention
• Added finer grained locks for pool and index access to improve SMP performance.
• Optimized index retrieval to return sorted choices
• Added tools to manage malformed UTF-8 text (important for DMOZ integration)
• Improved ability to preload indices to improve performance
• Fixes to HTML/XML integration with special attention to XML
• Fix mime parsing with BASE64 encoding
• Added tools for parsing structured (tab, space, etc) plaintext
• Separated out the FDTEXT and FDWWW libraries
• Made FDXML actually be usable and added callout models to have FDXML call Scheme to generate content.