Modern Software Experience

2008-05-13

Geni.com

social genealogy

Geni is one of the many new social genealogy web sites. Geni does not just let you build a tree, but actively encourages you to invite the family you add to become members themselves. As you add people to your tree, you can invite them to help you build that tree. This makes Geni.com a viral website.

web site

Many have remarked on the strong visual appeal of Geni.com home page. Everyone seems to love the visual design so much that they entire fail to notice serious technical design issues.

Geni is new and unburdened by legacy code, yet their web site does not validate. Even worse, Geni does not even show any menu until you enable scripting! Once you hesitatingly allow scripting, Geni shows it menu along the bottom of the screen instead of the top.

changed

If you followed Geni from the start, they probably struck yet another bunch of American guys with little knowledge of genealogy, who were convinced they were about the change the world, and what happened next is that the world changed them.

They became aware that the web is a world-wide web. They became aware that users in other countries use different character and have other keyboard layouts.

If there is one thing Geni.com has done right, it is listening to their users. They added privacy features. Geni now displays maiden names. They even added timelines - a feature many desktop programs still lack.

signing up

To use Geni, you need to create an account first, and creating an account with Geni feels slightly different from creating an account with an average website. It is just the same, you provide your name and email address, but you do so on their home page, in the diagram that show you as the child of Your Father and Your Mother. That does not really mean anything, it is just a cute design that Geni.com uses to use a sign-up form as their home page. The only difference with other sign-up forms is that Geni uses the name you enter to create a tree with just you in it.

By the way, if you signed up before, but forgot your detail, just try to sign up again, and Geni will send you an email with a link to reset your password.

Geni.com is quite aggressive about logging out non-active users. If you don’t do anything for perhaps half an hour, you cannot continue your sessions, but need to log in. That is annoying.

flash

Geni biggest design issue is probably the use of Flash for the user interface. Flash makes is relatively easy to make something good-looking, but not so easy to make it perform good. To be blunt: Flash does not derive its name from that comic book character. Trying to use Geni over a dial-up connection must be a test of patience.
If there is a trick to making Flash web sites perform, Geni does not seem to have mastered that trick yet. When I select tree from the main menu, it takes Geni about five seconds to display a tree with one person in it. I can easily download the entire Flash player in the same time.

features

I had a brief look at some of Geni.com current features. The awkwardly-named Cycle feature is evidence that the Geni team is not entirely in touch with genealogical thinking yet. And that profile pages still expect phone number in US-centric format is precludingly parochial.

It is not just tree display that is slow. Searching through all trees for a particular name is slow; Geni.com is not just slow to give a first screen, but even displaying the next page of search results takes ten seconds (that is not an estimate, I clocked it).

GEDCOM

Beta

The Geni people have been working on GEDCOM support for about a year now. When they first started on this, the discussions on their forum made it painfully clear that they had never heard of GEDCOM before and had no idea what they were doing. There was quite a lot wrong with the initial support, which they released in "Alpha". It was publicly available to all user of the site, so it really was a Beta, but it was certainly Alpha quality.
By the way, if the LDS needed any real-life proof that the numbering they use for the GEDCOM specification is messed up, Geni.com initially thought it a good idea to support GEDCOM 6. Luckily, they now support GEDCOM 5.5.

Release

A few days ago, the Geni folks decided that their GEDCOM support is finished. I am sorry to say that a quick look at their wiki documentation makes me doubt that. They write "Sometimes foreign language characters don’t seem to export properly. Try opening the file and changing the line that reads CHAR UTF-8 to CHAR ANSEL or CHAR UNICODE and see if that resolves the issue.".

I cannot have much confidence in a company that displays such a lackadaisical attitude towards your data. If you follow their erroneous advice, you will be messing up your text by making sure that all non-ASCII characters are interpreted wrongly.

The wiki allows registered users to contribute content, but this particular line was added on 2006 Nov 16 by Scott Steadman, who seems to be the Geni.com employee responsible for implementing the GEDCOM support.

tag list

I do have to applaud Geni for doing something I wish all creators of genealogy software did; they provide a list of the tags they do and do not support. A quick look over that list suggests that their GEDCOM support is fairly complete. That they do not support Ancestral File Numbers (AFN) is hardly an issue, as it is an LDS-specific GEDCOM feature. More relevant is that they do not support surnames prefixes (SPFX).

testing

Geni does not merge GEDCOM files into existing tree. When you upload a GEDCOM file, Geni will always create a new tree from it. That is ideal for testing imports, but not so ideal is that the import screen asks me to provide my first and last name as they appear in the GEDCOM file.
There is a mismatch between what I want to try and what Geni wants me to do. Most desktop software and websites do not care what you research, Geni does. Geni specifically wants me to upload my own tree, while I want to upload some other research.
You might think the way around this is too just picking some name from the file, but I fear that Geni will immediately start making connections to other users, and that it will be impossible to unload the tree.

I therefore decide to start my import test with a small nonsense tree I once created to test support for some particular characters. When I do so and click next, I get to see a screen that tells me my email address has already been taken. Perhaps the import screen the Geni blog links to is not the one I should be using. Or perhaps the system is confused. It is displaying a login button although I am already logged in. I try logging in again, and that is successful, but it brings me to my home screen, not a more promising looking import screen.

no import into your account

There seems to be no way to import a GEDCOM while logged in or using the email address you already registered with. That’s weird. Surely, after more than a year of Beta, that should work right?

No. As the blog announcement points out Importing a GEDCOM will create a new Geni Tree. If you already have a Geni account you must use a different email address for this import. That makes the import feature pretty lousy.

doubtful reasoning

There is a reason behind this madness, and it is not hard to see. Geni.com, perhaps somewhat prematurely, has just announced that the GEDCOM import feature is done, but has not announced a tree merge feature yet, and were you to import a GEDCOM into your existing account, you would surely get duplicates. As far as I can tell, that is the reasoning behind the demand to create a separate account.

Why not let everyone import an additional tree and then merge with their existing one? Doesn’t Geni already have logic to handle duplicates? Surely the whole point of sites like Geni, where we share our data, is that we discover ancestors in common with other users? No, a brief sojourn through the forums reveals that Geni does not offer the ability to merge with or link to another tree at all. The help has information about merging nodes or profiles, not trees. New users can join an existing tree and help grow it, but when two trees overlap, well, they overlap, and That’s it. Rather disappointing for a web site with the Everyone’s Related slogan as a by-line.

Still, at least the users who signed up, and never expanded their tree beyond themselves, but just waited for the GEDCOM import feature, should be given the option to import a GEDCOM.

GEDCOM larger than 5000 INDI

I decide to hit the forum and see what others are saying. The first thing I learn when reading through the comments is that the import feature is limited to GEDCOM files with at most 5000 persons. Shame on Geni.com for that not mentioning that in their announcement!

One users remarks that he has a 26K INDI GEDCOM containing his mother. He cannot import it because it contains more than 5000 INDI and because it does not contain himself. Another remarks that he signed up and then waited for the GEDCOM import feature, and is disappointed that he cannot import the data into his existing account.

queuing up

Some forum remarks by Scott reveal that Geni adds GEDCOM export jobs to a queue to be processed. I am guessing it is a first-in first-out (FIFO) queue, and that the overall user experience would improve if Geni were to prioritise jobs size to get small jobs done quickly and on age to make sure big jobs are not eternally bypassed in favour of small ones. A desirable side effect from such prioritisation is that the queue remains relatively small.

GEDCOM export

The GEDCOM link in the menu along the bottom of the window does not bring to a GEDCOM page, but to a GEDCOM export page.

Exporting my three-person tree is as easy as clicking the big "Export FirstName LastName’s GEDCOM file" button. There are no options to choose from, just click. Geni does not pop-up a File Save As dialog box. Geni does not allow you to save your tree to a file of your choosing, but emails the GEDCOM to your registered email address instead. That is less flexible, but also very simple to use, a lot easier for Geni to support and just a tad more privacy-conscious than letting any logged-in user download your tree.

The export speed is definitely disappointing. Geni tells you that the file is being created, and that is it. There is no progress bar and Geni does not indicate how long you will have to wait for the email to show up. Generating a GEDCOM for a three-person database should be near-instantaneous, but five minutes later, I still had not received any mail. Ten minutes later, I had still not received any email. A quarter of an hour later, I had still not received any email. Half an hour later, I had still not received any email. An hour later, I had still not received any email. Two hours later, it is there. Actual time is one hour and 56 minutes.

export speed

One hour and 56 minutes is 6.960 seconds. Thus, the measured export speed is less than 0,00043 persons per seconds. A three-person file is hardly representative, but if I were measuring the export speed of various systems, Geni would stand a good chance of ending up at the very bottom of the pile.

GEDCOM download

The email does not contain the GEDCOM file, but a link. That is a download link for the GEDCOM, but a link to a download page. That page has just a "Download My GEDCOM File" button. Unlike many other Geni features, this page does not demand JavaScript. The actual download file is not a GEDCOM file, but a zipped GEDCOM file, name export-geni.zip. Once the download is completed, the page informs that the file remains available until 2008 June 3, that is a three-week file retention.

GEDCOM quality

The GEDCOM export did not offer any options. There apparently are no choices but the Geni defaults, but these Geni default are good; the export-geni.ged file is an GEDCOM 5.5 file encoded in UTF-8. The file lacks a Byte Order Mark (BOM), and that is a pity, as a BOM not only ensures correct display in modern text editors, but also helps to prevent import into genealogy programs that still do not support UTF-8 and therefore cannot import the GEDCOM correctly.

The GEDCOM export speed is an embarrassing all-time low, but I was pleasantly surprised by the quality of the GEDCOM file itself. The three-person database is a ridiculously small and contains nothing but the names. It does not contain dates, places, notes, or sources, and is certainly not a thorough test of the GEDCOM export. Yet, I still could not help noticing that the output is the cleanest GEDCOM file I have seen from a one-dot-zero product.

The GEDCOM is nearly perfect, and identifies the source as Geni.com, version 1.0. I wonder whether the version number refers to the Geni.com site, which is still sporting a Beta tag, or just the GEDCOM export. The answer to that question does not really matter; as long as Geni.com keeps treating this field consistently. That the DEST is missing is just a small oversight. The file tag is missing, which is a bit odd, as Geni.com seems to be using a separate database for each tree. The SUBM tag is present and correctly refers to the creator of the database.

The GEDCOM file even has change dates (CHAN tag) on INDI records like desktop programs. But one for one silly exception, the GEDCOM is free of proprietary extensions and contains standard GEDCOM tags only. The only mistake I noticed is the use of EMAIL, which is not a GEDCOM 5.5 tag and illegal as a GEDCOM extension; GEDCOM extensions must start with an underscore. Geni.com should either export data to GEDCOM version 5.5.1, or opt to use a proprietary extension such as _EMAIL.
The biggest structural mistake I noticed is that Geni always includes a BIRT tag, even if there no birth information at all. The BIRT tag is optional and should not be included if there is no birth information to include in subordinate tags.

It is not perfect, but a not entirely unknown vendor of an infamous product that has already reached version 17 (a.k.a. "2008") but still does not support UTF-8 or write a valid GEDCOM header could learn something from these upstarts.

database deleting

I cannot upload a file without giving all the data to Geni.com. The Geni.com site offers the ability to delete individuals from your tree, but removes that ability as soon as they are a branch instead of a leaf on your tree. There is no way to delete more than one individual at a time. Now, most desktop software does not offer that ability either, but it does offer the ability to delete the entire database. Geni.com does not offer that ability.

import speed

As long as I cannot delete a database after making it, I will not perform an import on Geni.com. It 5000 INDI limitation means it cannot handle the 100k INDI GEDCOM anyway.

Some other members have uploaded data, and a forum message left by Scott Hibbard tells us that loading a file with 4976 individuals (just a few more than my 1 MB GEDCOM file) took two hours. That suggests that, everything else being equal, import 1 MB GEDCOM with 4862 INDI would take approximately 1 hour, 57 min and 15 seconds, which is 7035 seconds.

That is so slow that even a laggard like TMG 7 finishes importing the 100k INDI GEDCOM before Geni.com finishes importing the 1 GB GEDCOM. It is an import speed just above ⅔ INDI per second. It is considerably less than one individual per second and barely more than a measly 150 bytes per second. It is a new low that puts Geni.com 1.0 at the very bottom of the GEDCOM Import Speed pile. I had not thought it possible, but Geni.com manages to make Genea 1.4.1’s import speed look good.

By the way, if we assume an import speed of exactly 2 INDI per 3 seconds for the 100k INDI GEDCOM, import would take 150.100,5 seconds, which is 1 day, 17 hours, 41 minutes and 40,5 seconds. As performance is likely to degrade with GEDCOM size, an import time of about two days does not seem an unreasonable guesstimate.

conclusion

no merging

Perhaps the most important thing Geni.com does not exactly advertise is that they do not support merging of different tree. Their own "Everyone’s related" slogan explains what’s wrong with that. Geni.com does not even support Smart Match yet; you do not get to know when your tree overlaps another.

no database deletion

Another thing Geni.com conveniently forgets to tell new users is that they cannot delete their database. I like to remain in control, so I did not upload my test files. I therefore do not know whether Geni.com’s GEDCOM import supports ANSEL, how well it handles GEDCOM extensions or how good the import log is.

only small files

Another thing that slipped Geni.com’s mind when they let the world know that they happen to support GEDCOM import now, is that their shiny new GEDCOM import will only accept small files containing no more than 5.000 individuals. Slow GEDCOM import may be one reason for this serious limitation, but I am guessing that the less than spectacular performance of their Flash application is the more important consideration..
The inability to import GEDCOM files into your existing tree is a rather silly but highly annoying limitation to the current import process.

slow is an understatement

One thing Geni.com does not mention either is that calling their import process slow is an understatement that does not even begin to hint at the spectacular lack of performance. Estimated performance is considerably less than one individual per second. At the same speed, importing the 100k INDI GEDCOM would take days.

some good things

Their wiki documentation contains erroneous advice, but let’s end on a positive note. There are some good things to say about Geni.com GEDCOM support. Geni.com actually post a list of GEDCOM tags they support, and they seem to support practically all the tags that matter.

Moreover, I had a brief look at their GEDCOM export. It may be even slower than the GEDCOM import, and a simple three-person tree with nothing but names is hardly a test of its capabilities, but I was positively impressed by the simplicity of the export (no options), the optimal character encoding choice and a fairly clean GEDCOM file, including a near-perfect GEDCOM header.

updates

2008-06-19 GEDCOM Import limit raised

Following some performance enhancements, Geni.com has raised the GEDCOM import limit from 5.000 to a less unreasonable 15.000 INDI records.

2008-10-09 GEDCOM Import limit raised

Geni.com has raised the GEDCOM import limit from 15.000 to 50.000 INDI records.

2011-04-23 Geni.com knowledgebase How do I merge two family trees?

The Geni.com knowledgebase article How do I merge two family trees? (answer: you cannot) seems to have been deleted. It probably was seriously outdated, as Geni.com has not remained unchanged since 2008. The broken link has been removed.

Geni.com 1.0 estimated performance

file1 MB GEDCOM100k INDI GEDCOM
time1h57m15s-
time in seconds7035-
INDI per second0,69-
bytes per second150,09-

Links