Slashdot Mirror
Google Revises Usenet Search
michaelmalak writes "Wednesday night, Google Groups announced in a thread the rollout of their revised 20-year Usenet archive search engine. Among the various 'improvements': ability to search by date has been eliminated, as has the ability to deep link to a single post. See the announcement thread for others' reaction." An anonymous reader writes "ZDNet has published some interesting insights into what makes Google tick. In this lengthy article, Google's vice-president of engineering, Urs Hölzle delves into the nuts and bolts behind Google's operations, what back-up mechanisms and hardware setup is in place and even some interesting homegrown technology like the Google File System (GFS)."
The article states:
- Over four billion Web pages, each an average of 10KB, all fully indexed.
- Up to 2,000 PCs in a cluster.
- Over 30 clusters.
- One petabyte of data in a cluster -- so much that hard disk error rates of 10-15 begin to be a real issue.
- Sustained transfer rates of 2Gbps in a cluster.
- An expectation that two machines will fail every day in each of the larger clusters.
- No complete system failure since February 2000.
Now, 2,000 machines in a cluster, plus 1PB data, plus 2Gbps in a cluster times 30 clusters comes to:
- "Over" 60,000 PCs (!)
- "Over" 30PB data storage
- "Over" 60Gbps bandwidth
Also interesting:
- An expectation that two machines will fail every day in each of the larger clusters.
- No complete system failure since February 2000.
I'm believe that the "new groups" are not new usenet groups, but merely a yahoo-groups clone on the side, which gets he same interface as the one they provide for usenet groups.
The old groups interface rocked. This is a major step in the wrong direction in my book.
Unable to read configuration file '/bigassraid/htdig//conf/14229.conf'
Geocrawler error message.
You can still do a deep link to a single article, if you like....
Navigate to the thread, for example this comp.arch thread. Choose the post you want to link to, and click on "Show Options". Two of the options are "print", which is a link to a "printable" version of the article, and "Show original", which is a link to the article with all the headers.
One more step (or simple URL hack) from this display is "view parsed" which gives a friendly HTML version -- for example, try this link.
You are wrong. You are not on the new Google Groups page. There is sort by date, but not search by date. You want to look at groups-beta.google.com, not groups.google.com.
I was actually lucky enough to visit a datacenter in the southeast united states (which will remain nameless, but if you do a little searching, Im sure you could figure it out) where Google colocates. I want to say they had something like 18,000 square feet just for them, behind a partitioned wall. We were *not* allowed back there, despite my pleading.
Anyway, as we were walking around the 150,000+ square foot datacenter floor, when a guy came by, pushing a very odd looking rack.
It resembled a bread tray, 20 shelves if I counted correctly, with completely naked main boards sitting on them. It looked to be 4 machines per row (counting the power supplys). Each had one IDE disk sitting on a gel pad, strapped in with velcro. I personally watched them wheel 4 of these racks right by me back into the dark "Google" corner of the datacenter. Our tour guide finally gave in.
Him: "Well, you've seen them now!"
Me: "What do you mean?"
Him: "Thats google!"
Definitely the highlight of my day!
Who was the idiot that started this rumor?
Each message in a thread has a named HTML anchor, try this for instance. It will show the whole thread, but position you at an exact message in the middle.
The only problem is there is no easy way to get this URL, you have to find the anchor by looking at the HTML source (Firefox's "View Selection Source" feature helps a lot).
Also, if you click on the "Options" link by the individual message, you get a "Show original" link, which shows just the message, verbatim.
And from there, you can click on "View parsed", and see just the pretty message, without the rest of the thread.
So there's your deep-linking. I agree it's not obvious how to do it at the moment, but the ability is obviously still there. Give it some time, it's still a beta!
These quirks and the "Server Error" bugs are to be expected, they'll work it out.
As for the new browsing interface itself, I kinda like it. It integrates and borrows some stuff from their excellent Gmail interface.
It hides quoted text by default (you can expand it with single click), so you don't have to scroll through some morons quoting of a whole message just to add a few words, it keeps a history of groups you recently visited, it allows you to bookmark topics you are interested in, etc. I do find it an improvement over the old interface.
The only thing is the missing date search, I agree there, that was definitely useful feature. If enough people complain, maybe they'll bring it back.
Also, someone else complained that you cannot browse by group anymore... bullshit, it's staring you right in the face, it's the "Browse all of Usenet" link.
Alright people, you can stop overreacting. They just rearranged some things, that's all.
There's a link at the top of the thread to turn on the left-hand tree frame.
Deep-linking to a single post is still very much possible.
And I highly doubt that a search-by-date feature is going to go missing for long in a 20-year archive. This is, after all, a BETA.
As per usual, Slashdot editors didn't even think it worth their time to follow a single link to see if the submitter wasn't trolling.