EXCEPTION::QUERY

A Blog About Database Products and Technology.

September 2006

FourJs Software makes Genero, a multi-platform, "write once, deploy everywhere" business-application development environment comprising a complete IDE and a SQL-aware 5GL programming language with simple connectors to front-end and client environments and web services, that plugs and plays with most popular SQL DBs. They market it as a solution in situations where OOP vs. Database or skills vs. platform-deployment needs are becoming unmanageable, and it is extremely slick. They also make generoDB, a high-capacity, high-performance SQL database optimized for "robust realtime" projects. It performs more like an in-memory database without sacrificing ACID expectations. Has automated failover and backup, replication -- the whole nine yards.

And guess what! We have all four versions of generoDB (for Linux32, Linux64, Solaris and Windows) available for free download and trial. Hustle on over to our little registration page, surrender some easy-to-replace personal data, and grab a version for your favorite OS.

generoDB is architected to minimize locking and promote concurrency, while preserving data integrity. They accomplish this by efficient use of large amounts of RAM, under control of a concurrency engine and multi-platform, multi-threaded dispatcher; and by realtime JIT compilation of SQL, among other strategies. In enterprise client/server and collocated deployments, generoDB lets you get away with fewer (and older) servers for a given performance level (they claim they can save you between 40% and 60% on hardware). But generoDB is also set up to exploit heavier-than-usual iron (e.g., huge RAM complements, symmetric multiprocessing), and can be deployed in an embedded-scale system factor as well. Robust SQL version and high-level language compatibility means it can play well with existing applications, and avoid confining future ones.

Posted by John Jainschigg at 05:11 AM  Permalink |

At SD Best Practices in Boston, this morning, Wizart (www.wizart.com) showed me a crisp little product for localizing finished Windows and web applications. There's no machine translation involved -- the core of the product identifies where text (e.g., prompts, user instructions, button labels, etc.) appears, based on its instrumentation of Windows or interpretation of XML, HTML or other web UI representation), and provides a console (WizTom) for translators that lists these items and lets the translator input contextually-appropriate equivalents. These are then maintained in a table accessed by a coresident process (in Windows) or server proxy (for web stuff) and inserted in running applications on the fly. Result: a translator works for a while, and boom, your application is localized.

WizTom supports the Windows default localization library for common menu items and terms (e.g., File, Edit, etc.), so you may not even need a translator (or even access to source code) in some cases. The WizArt folks demoed this by converting Windows NotePad, on the fly, into French, German, Chinese and Arabic. Source code mods are generally unnecessary unless you're retrofitting, for example, an application not intrinsically Unicode compatible.

Very nifty feature in the Windows version: when viewed under control of the translator/developer console, resource properties relevant to localization are given handles and can be manipulated directly (e.g., to expand buttons so their Chinese labels fit). And OS-level localization options are engaged automatically (e.g., flipping editable text direction and scrollabars from left-to-right to right-to-left when translating into Arabic or Hebrew). This data becomes part of the coresident set and governs the appearance of the app in its final, translated form.

Not cheap! But if you need it, much cheaper/faster/better than other alternatives.

Posted by John Jainschigg at 12:24 PM  Permalink |

Announced Sunday (which is, one suspects, the 'new Friday' for public disclosure of controversial news), the FBI and the Department of Homeland Security have finished merging two large-scale fingerprint databases, and are now rolling out front ends to police departments, FBI investigators and DHS and Immigration agents, airport inspectors, border agents and other authorities. The contributory databases are IDENT, a DHS database containing over 1 million fingerprint records of illegal immigrants ordered deported, and the FBI's IAFIS (Integrated Automated Fingerprint Identification System), containing biometric data and criminal history on 47 million people.

The fused system will give FBI and DHS investigators immediate access to all pooled data -- improving greatly on an unweildy, non-realtime query-and-reporting system in place before.

According to Robert Mocny, Acting Director of US-VISIT (the agency tasked with managing the DHS database), as quoted in an article by Marisa Taylor of McClatchy Newspapers (reprinted yesterday by the San Jose Mercury News), the Dallas and Boston Police Departments will be among the first to trial the new system -- and Mocny intends to make it generally available to police departments as scaling issues are worked out.

Though not explicitly stated, Mocny implies that one scaling goal will be to increase capacity and streamline input process for the IDENT database, so maintenance doesn't overwhelm immigration officials. Taylor's article quotes him as saying: "We've got about 11 million illegal immigrants in the United States ... We want to build up the capacity of how we're going to be able to respond."

Though legal bars doubtless exist governing the ability of police departments to add records to the central store; and though there's been no ostensible alteration of IDENT's brief (i.e., to contain only records of immigrants under order of deportation after due process), the potential is pretty clear -- the mid-term goal is one, big database referencing all illegals intersecting with police; the longer-term goal, potentially, to extend this database to include all visitors of questionable status.

Even in the short term, access to this database removes a substantial practical bar to police departments enforcing immigration laws -- something police (and most local governments) are not, by and large, itching to do, since they believe it will conflict with practical law enforcement, community relations, and business. It will be interesting to see whether DHS and the administration can mobilize to begin forcing the issue.

Meanwhile, the IDENT component of the database may not be 100% useful, since it currently contains only two fingerprint impressions per record -- insufficient data for a positive ID in some cases. US-VISIT plans to upgrade to a full ten-impression set over the next two years.

Posted by John Jainschigg at 11:03 AM  Permalink |