Searching a Windows system has always been harder than it should be. Windows has come with an indexing service for a while to facilitate searching, but it and the search are second-rate. Google must have thought it a natural to bring its highly regarded search technology to the desktop.
The company also should have counted on the negative publicity it got, even though all of it, as best as I can tell, is based on bad analysis—some of it from suspicious sources. I looked at the program and the complaints, and I dont see a big deal here.
The just-released beta version of Google Desktop Search is an attempt to make a local search engine with the power and charming interface of the popular search engine. It indexes and then lets the user search e-mails, chats, Web sites the user has viewed and files on the users system.
I installed GDS on two systems. Its a svelte download (456,808 bytes) and fairly small in memory as well: three processes, totaling about 12.5MB RAM. (Sigh … How our standards have changed. 12.5MB used to be a lot of memory.) The search index, stored in the C:Documents and Settingsuser nameLocal SettingsApplication DataGoogleGoogle Desktop Search directory, took up 757,002,240 bytes.
This is definitely a lot, but not for this particular system. My biggest complaint about GDS while its running on my system is that it adds yet another tray icon. Windows XP started to deal with the problem of tray clutter, but its still largely out of control. But back to the issue at hand.
After the program installs, it begins a one-time indexing process that, depending on your configuration, can take some time. On one of my systems, it took several hours because my My Documents folder is located on a network drive (a simple change to make in Windows XP). This shows, by the way, that GDS doesnt just index the local drive.
The search facility worked very well for me and is extremely fast. It runs as a Windows service, accessible through a local http server on a nonstandard port. The absence of any user interface code, other than the HTML generation in the engine, must be partly responsible for the small size of the program. The GDS page itself is a classic, simple, Google-style Web page served from http://127.0.0.1:4664/ with some indecipherable options on the URL.
The results appear as a Google page, including e-mail results. As a slick bonus, entries for Web pages found in the cache also have a thumbnail image of the Web page itself.
The results for Outlook e-mail searches are also slick, although not as convenient as some alternatives. I use a different product for Outlook called Lookout by Lookout Software, recently acquired by Microsoft.
Lookout works reasonably well, but is Outlook-only. On the other hand, the search facility works within Outlook using Outlook facilities. So, the results list is a sortable list of messages; you can sort on the folder, date, etc. GDS results are a Google-style browser list.
Dont get me wrong, GDS looks like it does everything it could do from a browser, more than youd think possible. When you view a message, it includes links to view it in Outlook, reply or reply to all. Most of these links take you right into context in Outlook itself. I presume it works similarly in Outlook Express, but I didnt test that. I use Outlook 2003.
GDS lacks some other features that Id expect as Google refines it, such as the ability to search within the result set. The sorting facilities arent as slick as Lookouts (or some other products), but there is the ability to sort results by date as opposed to the default “relevance” (as determined by Google),and you can view just files or e-mails or Web history or chats.
So, GDS isnt quite as powerful on second glance as it is on first.
Privacy and Mixed Search
Results”> Exhibit A for Google Desktop Search critics that its a real tinfoil-hat job: As explained in the GDS documentation, you can combine search results from the desktop and Google Web search service, and it happens automatically. Once you install GDS, if you do a Google Web search, there will be an entry in the results linking to your local results.
On the one hand, it seems obvious to me how this is being done. As I said before, because the installation warned me, GDS installed an Internet Explorer add-on. I suspect it is involved with this result. When it detects a Google search, it calls the local GDS engine to do a search and inserts its results.
I dont see any privacy problems with this because all of the local results stay local. David Burns, CEO of Copernic, disagrees. In a story in The Register, he states that this will result in Google knowing whats on your computer.
For some reason, the CEO of Google competitor Copernic sees things differently.
The terms and conditions state that GDS will collect non-PII (personally identifiable information) such as the number of searches you perform. This will be sent to Google unless you opt out. It will be used to improve the program and will not be sent to any third party, the company says.
Theres a unique application number in each instance of the program. Google gets this number back at install time, along with a code indicating success of installation. Its also sent when the program checks for a new version. All done to “make the software work better.” Sounds to me like they want to know how many copies there are out there, but they dont know whos running them.
As previously mentioned, it uses the central Google cookie. They may use this along with the information about other Google services (such as your news and Web searches). You can opt out of this anytime.
You can stop the system from indexing any particular file or files, or remove such files from the index after they have already been created.
On the other hand, Google seems to think theres something to hide with the desktop searches. When you do one of these mixed searches, the desktop results have a small “Hide” link next to them which, when clicked, causes the desktop results to disappear from the page. The documentation says, “You may want to hide a particular set of your personal Google Desktop Search results if someone is looking over your shoulder or if youre projecting to an audience.”
And of course, you can always stop the mixed results from ever showing up, or stop specific files and folders from being indexed (go to Preferences on the GDS home page; go to the “Dont Search These Items” section). So, if youre actually concerned about search results showing up, there are things you can do.
Privacy and Multiuser Systems
The second main complaint Ive seen is that GDS searches the entire local system, including the files for all users, not just the current user or the one who installed. Theres something to this, in that it means that one must use GDS with special care—if at all—on systems with multiple users.
The documentation makes clear that GDS is intended for a single user machine, or at least that it can only be installed on a single users login for a particular computer. Exactly how it functions on a multiuser system depends on how that system is configured and who the user is who installed GDS.
But its already been pointed out by others that since GDS indexes the entire computer, it will index the private files of other users on the system. You may, therefore, view files belonging to other users. If they use Web mail, you may be able to view their mail.
Lets think this through. GDS will only install on a system when the user doing the installing is an administrator, and it will only run when the user who installed it is logged in. I tested this myself.
So, in order for you to be able to view the files and browser caches of other users on the system, you must be the system administrator. Guess what, Sherlock! You already had access to those files! Shocking as it may be, GDS only makes it easier to do what you already had explicit power to do.
This objection is speciousness defined. If you still dont like the prospect of the system administrator having access to such information, you can mitigate it somewhat by modifying GDS options to have it not index https pages, which will eliminate most Web mail. You also can configure Internet Explorer to delete all temporary files when it quits.
Platform and Browser Restrictions
Google has taken an unsurprising amount of flack over its limited platform support for Google Desktop Search. GDS help files and FAQs have many apologies for their limited support for applications and platforms. They promise they are considering support for others: an emphatic “maybe.”
GDS only works on Windows XP and Windows 2000 with Service Pack 3 or higher. This is partly because it works as a Windows service and partly because Google only wants to use recent, supported versions of the OS. It only searches Web caches of Internet Explorer from IE5 or greater (although no recent copy of Windows 2000 or Windows XP would use anything less than IE6).
It only searches e-mails from Outlook 2000, 2002 and 2003; or Outlook Express 5 or greater. It searches files in several standard formats such as TXT and HTML, but also Microsoft Office formats such as .DOC, .XLS and .PPT for Office 2000 or greater. It also says it searches chats from AOL 7+ and AOL Instant Messenger 5+, but I didnt test this, and the utility of it escapes me at first glance.
So, you may notice the absence of Mozilla cache support, OpenOffice formats and other things such as Linux support. I hope were all adult enough to recognize the fact that if one is prioritizing formats to support, IE and MS Office obviously are more important than Mozilla and OpenOffice. Dont count on expansion to the more obscure formats or other operating systems anytime soon.
Google already supports many of the same Microsoft formats on Web searches (see my column on privacy implications of this feature). Its an interesting mix of file formats that it supports using the filetype operator, and one would think the software might be the basis for file support in GDS, but not necessarily.
Software on google.com operates under resource constraints very different from those of a service running on the average desktop. It may be that supporting all of these formats isnt worth the cost in terms of memory. And none of the open-source formats are up there.
One last bit of criticism struck me differently than its author. Mike Langberg of the San Jose Mercury News thinks the AIM logging is a potential privacy problem because AIM sessions have always seemed so transient, and GDS makes them persistent. The same could be said for searches of e-mail and Web pages; you think theyre gone, but with GDS theyre not.
All I can say is “duh!” The whole point of GDS is to make these items findable. If they were still in the forefront of your memory, you wouldnt need GDS. One might as well criticize airplanes for making far-off lands too easy to reach.
I was concerned about the impact on system performance for a service that has to index files and Web pages periodically, but I havent noticed anything yet. Google claims that GDS builds indexes only at idle time, so it shouldnt affect system performance. Of course, lots of programs run only at idle time, and there are systems out there so busy that they have little or no idle time, so there are cases where it will affect performance.
But these should be increasingly rare over time since hardware continues to outstrip software in terms of performance (especially for a product that requires XP or Win2KSP3+). If your system is so overloaded that it has no idle time (check Task Manager), you should probably get yourself some more memory.
I was rough on Google competitor Copernic before, but I should point out that I havent tested its products, and they may well be much better than GDS. Ive already pointed out that in many ways the Outlook searching is inferior to Lookouts. Its just a beta, of course, but consider that Google News has been in beta for a long, long time. We should take this version seriously.
GDS is most appropriate for home, single-user systems. My own environment is more of a business network and it works well here, but only when I run it as an administrator. As things currently stand, GDS has a big problem with business environments in that it must be installed as an administrator and will run only in the context of the user who installed it, and therefore it must be run as an administrator. No business should allow this. Home users shouldnt really do it either, but they often have to.
This user-restriction problem is GDSs biggest flaw. Long before Google spends time supporting Linux or OpenOffice formats, it needs to clear this problem up.
Security Center Editor Larry Seltzer has worked in and written about the computer industry since 1983.
More from Larry Seltzer