Ben Langhinrichs

Photograph of Ben Langhinrichs
E-mail address - Ben Langhinrichs






October, 2017
SMTWTFS
01 02 03 04 05 06 07
08 09 10 11 12 13 14
15 16 17 18 19 20 21
22 23 24 25 26 27 28
29 30 31

Search the weblog





























Genii Weblog


Civility in critiquing the ideas of others is no vice. Rudeness in defending your own ideas is no virtue.


Tue 3 Oct 2017, 05:55 PM
Inline JPEG image
 
Last night, we posted CoexLinks Fidelity 4.10, which includes our new Defect Detection subsystem to correct flaws in images and attachment metadata and more. This is also the first public release with support for 64-bit Linux along with 32-bit Linux and both 32-bit and 64-bit Windows. All functionality is equivalent between the versions. Customers receiving earlier versions of 64-bit Linux for 4.00 and 4.01 should upgrade when feasible.
 
Other areas of improvement:
 
1) Performance improvement, approximately 20% depending on mix of email
2) Enhanced adherence to strict CSS/HTML5 guidelines
3) Enhanced invitation/event fidelity
4) Improved fidelity for buttons and some overlaid hotspots
 
Several outlier bugs were identified and fixed, though most customers have received at least a subset of those in interim releases. This version includes all identified bug fixes, including those found internally that were not identified in the wild.
 
Request an evaluation license and see for yourself how much more professional your mail can appear to customers, coworkers and management, whether it is read in Notes, another email system ssuch as Outlook, or on your smartphone or Verse.
 
Or if you want to see the results without installing anything, fill out the form below. We won't spam you or sell your address if you give us your email. It is only used for this demo and a quick follow up afterwards to make sure your questions have been answered.
 


 

Copyright © 2017 Genii Software Ltd.

Technorati tags:

Sat 16 Sep 2017, 02:24 PM
I wanted to demonstrate what our new Defect Detection (and correction) looks like, but the various examples I have are all covered under NDA. So, in this series, I will break some of my own images in the way the customer images are broken. (These are all scaled down from the original width to make them fit on the blog better.)
 
Original image (scaled down)
Inline JPEG image
 
 
1) Image header lists n segments, but there are only m (where m < n)
 
For my first example (going in the order of this blog post), I take the same image saved as GIF and JPEG (using GIMP) and imported into Notes rich text, then broken using the Midas LSX. In the GIF format there are two "image segments" in rich text, while In the JPEG format  there are six "image segments". This simply means that the raw image data is spread out over a few different CD records. When an image "breaks", it is often because some of the trailing image segments are lost, or rather overwritten. Below, I show the same image, first broken by removing the last few image segments (half of them), then fixed by the Defect Detection system in CoexLinks and AppsFidelity.
 
As you can see, many "fixed" images aren't perfect, but they are often legible enough to read. If the information on that image is critical, wouldn't you rather see of the those rather than the white box below?
 
 
Broken images (all look like this in Notes)
 
Inline JPEG image
 
 
Fixed JPEG image (minus half the image segments)
 
Inline JPEG image
 
 
Fixed non-interlaced GIF (minus half the image segments)
 
Inline JPEG image
 
Fixed interlaced GIF (minus half the image segments)
 
Inline JPEG image

Copyright © 2017 Genii Software Ltd.

Technorati tags:

Tue 22 Aug 2017, 11:41 AM
I got a few questions from people at MWLUG about what CoexLinks Fidelity is, and what exactly 'email fidelity' might look like. Rather than just talk, you can try for yourself. Note: our newer CoexLinks Migrate and CoexLinks Journal products use the same rendering engine internally. If you submit your email address below, you will get twelve messages to the email address you used. Six will be sent using the normal Domino 9.0.1 email engine, then the same six will be sent rendered by CoexLinks Fidelity. Compare and decide whether your company and your clients would be well served with email fidelity, or whether your migration would be safer with it. If you want to try CoexLinks Fidelity for yourself, just fill out an evaluation request, and we'll get you set up. To find out about the other two products, click on the links for each product above.
 
No, we won't spam you endlessly if you give us your email. It is only used for this demo and a quick follow up afterwards to make sure your questions have been answered.
 


Copyright © 2017 Genii Software Ltd.

Technorati tags:

Wed 16 Aug 2017, 05:01 PM
As part of our Defect Detection feature, we encounter and usually fix a number of different defects in images. The following list is the different scenarios we detect which cause any issue with extracting or rendering the images. Of these 31, at least 27 have been encountered in actual customer documents. The other four are left in because they might be some day. While none of these problems are common, some are likely to be encountered in any large data repository, especially in mail which has been converted back and forth between MIME and rich text (as with replies or forwards).
 
A few of these problems are completely unfixable, but only a few. For example, with scenarios 1a and 17, we have no image information to work with at all.
 
Many of these are easily fixed. For example, 25 or 26 are easily deduced from the data.
 
The rest can be partially fixed. Part of the image may have low resolution, or a section may be missing, but usually the image is at least intact enough to identify.
 
1) Image header lists n segments, but there are only m (where m < n)
1a)Image header lists n segments, but there are none
2) Image header lists n segments, but there are m (where m > n)
3) Image segments contain multiple GIF starts (often associated with #2)
4) Image segments occur with no image header
5) Multiple image headers
6) Image header lists data size other than actual (except where PNG)
7) Win meta header lists n segments, but there are only m (where m < n)
8) Win meta lists n segments, but there are m (where m > n)
9) Win meta segments occur with no Win meta header
10) Multiple win meta headers
11) Win meta header lists data size other than actual
12) Bitmap header lists n segments, but there are only m (where m < n)
13) Bitmap lists n segments, but there are m (where m > n)
14) Bitmap segments occur with no Bitmap header
15) Multiple bitmap headers
16) Bitmap header lists data size other than actual
17) No headers or segments
18) Both image segment or win meta and bitmap
19) Image resource that is missing
20) Pseudo-attached image that is missing.
21) Image segments occur before image header
22) Win meta segments occur before win meta header
23) Bitmap segments occur before bitmap header
24) No graphic record but other graphic elements
25) Graphic record says resized, but has 0 width or height
26) PNG record missing
27) PNG header lists n segments, but there are only m (where m < n)
28) PNG header lists n segments, but there are m (where m > n)
29) Multiple PNG headers
30) Image segments contain multiple PNG starts (often associated with #28)
31) PNG header lists data size other than actual
 
Phew! As you can see, there are a lot of potential issues, but with Defect Detection, we can fix the majority. That's why we do it.
 

Copyright © 2017 Genii Software Ltd.

Tue 15 Aug 2017, 12:53 PM
CoexLinks family of products: CoexLinks Fidelity, CoexLinks Migrate and CoexLinks Journal
 
Very soon, we are releasing a new version of all three of our CoexLinks products, CoexLinks FidelityCoexLinks Migrate and CoexLinks Journal. Aside from other features and bug fixes, they will share a new feature called Defect Detection. While the challenge for most document rendering (to MIME in this case) is faithfully reproducing the content of the email and including the envelope information in the desired form, some Notes emails have corruptions and defects which make the job harder.
 
There are four major defects (and a few smaller ones not worth mentioning):

  • Broken inline images. A variety of corruptions in images including zero-length data, missing image segments and incorrect image type (e.g., a GIF is marked as a JPEG) leave images broken in both the Notes client and the rendered document. We are able to detect and repair or partially repair about 75% of these corrupted images.
  • Compressed attachments with incorrect sizes. These are difficcult to detect because you can open the attachment or save it to disk from the Notes client, so you don't know you have an issue. But since the uncompressed size is incorrect, the document will be truncated and corrupted when emailed or when it is rendered by most tools including the Domino rendering engine. We can fix 100% of these corruptions.
  • Hotspots with invalid ends. In some versions of Notes, URL hotspots and other hotspots inside sections or table cells were left without a closing record. While they appear fine in Notes, they render with either large parts of the Body content missing, or with everything to the end showing as a URL link. We can fix about 95% of these corruptions.
  • Invalid stored image URLs. These corruptions are an artifact of the external MIME to internal MIME rendering, so mostly appear with received MIME emails or forwarded/replied to MIME emails. The fix is fairly simply, so we can fix 100% of these corruptions.
 
Whether you are sending email to customers, reading your own mail from a mobile or web interface, migrating an entire database or journaling mail to a third party vault, it is better to have defect detection in place so that unusual does not become the irretrievable.
 

Copyright © 2017 Genii Software Ltd.

Technorati tags:

Fri 21 Jul 2017, 03:56 PM
Most IBM Notes/Domino customers who have used the product for a number of years have vast stores of data, but when they want to try to glean new insights, they are stymied by how to handle the data mining. Simple fields which map well to views are easy to extract, and are often relatively "clean", meaning that the value is what the value says it should be. But real applications, especially those built for internal use, often reflect a far more complex set of relationships. They may use parent-child hierarchies, doclinks, lookups to other databases. They may also contain information stored in multi-value fields or rich text fields that require manipulation and cleanup. 
 
While there are a number of techniques available from DXL to data scraping, it can quickly become programming intensive to find information and put it together. With this in mind, we have built a fairly easy database using the Midas LSX engine to extract, correlate and prepare data from different sources and build a result which does not always have a one-to-one correspondence with Notes documents. The main virtue of this approach is the ease with which you can ask questions and put together sources. If you decide you have something wrong or need something else, it takes just a minute to remove or add it.
 
I wanted to show how this works with an existing application used over a period of years by fairly sophisticated Notes users. I chose as a source the IBM Business Partner forums, because they are  widely available and familiar. One of the different uses for these forums over several years was to allow partners to file Possible Bug reports, which IBMers could monitor and use to create SPRs and so forth. In this brief video, I pose five questions of this fairly simple application. Imagine how you could use a similar application to delve into your company's data. 
 
 
Note that I don't talk much in the video about data cleaning, but if you look at the image below, note that the column F (first red arrow) is derived automatically by Midas as a boolean from column G (second red arrow). We have some data cleaning built in as options, but are also looking at ways to provide custom data cleaning and normalization for individual items. While it is inevitable that some data cleaning will be done after the data is loaded into data analysis or data visualization software, the cleaner it can be the better, as 80% of all time doing data analytics is preparing and cleaning and normalization the data. We are eager to discuss with customers how we can minimize that costly effort.
 
 

Copyright © 2017 Genii Software Ltd.

Technorati tags: