Industries / Construction and property / Commercial and residential construction / Glazing
Sealing and testing
Sealing and testing is the work of keeping water and air out at the glass and around the frame, and of proving it with laboratory and site tests, ratings and labels.
6
Milestones achieved, from 1963 to 2024.
1
To watch for: announced, not yet confirmed outside.
2
Open deliverables. Nobody has shipped them.
The road
AchievedWatch forOpen
Achieved
1963
Rain kept out by design
G. Kirby Garden of the National Research Council Canada's Division of Building Research published Canadian Building Digest 40, "Rain Penetration and its Control", in April 1963. He named four forces that carry rain into walls and joints: the momentum of the drop, capillary suction, gravity and air pressure differences. A joint sealed with gaskets or sealant where rain can wet it "must be perfect", he wrote, and site errors, aging sealant and building movement make that hard to achieve and harder to keep. His answer was an outer "rain screen" with an air space behind it, open enough to the outside that its pressure matches the wind's, so no pressure difference is left to push water in. In a commentary published by RDH Building Science, John Straube calls it a classic, and adds that research in the 1980s and 1990s found full pressure equalization rarely achieved in the field: average pressures equalize, but gusts lasting a few seconds do not.
LessonA seal on the wet face of a building will open sooner or later, so a design that expects some water and lets it out lasts longer than one that relies on the seal.
Credit
G. Kirby Garden, National Research Council Canada
1977
Safety glass in doors
The US Consumer Product Safety Commission (CPSC) issued its safety standard for architectural glazing, 16 CFR 1201, on 6 January 1977. It covers glass in doors, storm doors, sliding patio doors and bath and shower enclosures. The Commission estimated that 73,000 injuries from glazing in such products were treated in hospital emergency rooms in 1975, about 2,400 of them serious enough for admission, and that children aged 14 and under suffered about half of them while making up less than 30% of the population. People walked or ran into glass doors they thought were open, fell through them, or were cut installing and replacing the glass. The standard set impact tests of 150 and 400 foot-pounds (about 200 and 540 J), depending on the product's category, and the Commission expected its main effect to be the substitution of safer glazing for ordinary annealed glass; wired glass, which could not pass the 400 foot-pound test, kept a place only in fire doors.
Credit
US Consumer Product Safety Commission
1992
A law backs one label
The National Fenestration Rating Council (NFRC), a voluntary coalition of window makers, utilities, government and consumer groups, was formed in 1989 to rate the energy performance of windows "in a fair, accurate and credible manner". Section 121 of the US Energy Policy Act of 1992 told the Department of Energy (DOE) to support a voluntary national rating and labeling program for windows, developed by NFRC, and to write a mandatory one if NFRC's fell short. In September 1994 DOE found NFRC's program acceptable, on conditions that included finishing a basic rating program by 24 October 1995. By then the California Energy Commission had made the NFRC program mandatory in California, and NFRC's procedure for rating heat loss (the U-value) had been validated by US and Canadian laboratories.
Credit
NFRC; US Congress; US Department of Energy
1994
Lumber fired at windows
Before Hurricane Andrew struck south Florida in August 1992, the region's building code did not require shutters to protect glass. Miami-Dade and Broward counties then wrote a new South Florida Building Code, the 1994 edition, whose largest change was that the building envelope must resist wind-borne debris. In its large-missile test, a 9 lb (4.1 kg) wooden 2-by-4 about 8 ft (2.4 m) long is fired at a window, shutter or wall at 34 mph (50 ft/s, or 15 m/s); after the impacts the specimen must withstand more than three hours of cyclic pressure like hurricane gusts, and three specimens must pass. Palm Beach County adopted the impact tests in 1995, Monroe County in 1997, and the first statewide Florida Building Code (2001) applied them in every wind-borne debris region. After Hurricane Wilma in 2005, when broken outer panes of some systems became debris themselves, the code required that outer pane to be safety glass.
LessonA broken window lets wind into the building and raises the pressure on its roof and walls, so protecting the glass protects the whole structure.
Credit
Miami-Dade and Broward counties
1999
Most fail their first test
In ASTM E1105, the standard field test, an installed window is sprayed with water while a chamber holds the air pressure outside higher than inside. P. E. Beers and W. D. Smith analyzed five years of such tests on windows and sliding glass doors in new construction and found that a majority failed their first field test, which they took as the reason to test during construction. A 2022 article in IIBEC Interface, the journal of the building enclosure consultants' institute, says reliable failure rates still do not exist because few projects test every window, and that testing 1% to 2% of installed windows is common.
LessonWindows that pass in the laboratory often leak once installed, and only a test on site finds out.
Credit
P. E. Beers and W. D. Smith
2024
Hurricane rain outruns lab tests
Under the North American standard, most windows and doors are water-tested at 15% of their structural design pressure (20% for the heaviest class), with a minimum of 2.86 psf and a maximum of 12 psf. By the window makers' association AAMA's own table, a window rated for 50 psf (2.4 kPa) is water-tested at 7.5 psf (360 Pa), the pressure of a wind of about 49 mph (79 km/h), and AAMA warns that the rain of tropical storms and hurricanes can exceed a product's water rating. A literature review for the Florida Building Commission, ordered by a 2024 state law on water entering through the tracks of sliding glass doors and submitted on 15 November 2024 by researchers at Florida International University and the University of Florida, concluded that the 15% protocols in Miami-Dade's TAS 202 and the AAMA standards "may not adequately assess" rain intrusion in real hurricanes.
LessonA window's water rating describes a moderate storm, not the hurricane its structural rating is meant for.
Credit
Omar Nofal, Arindam Chowdhury, Ryan Catarelli and colleagues; AAMA
Watch for
Canada drafts one U-factor
The first draft of Version 6.0 of Canada's ENERGY STAR specification for windows, doors and skylights, dated July 2026, sets a single maximum U-factor (the rate of heat loss through the product) of 1.05 W/m²·K (0.18 Btu/h·ft²·°F) for windows and doors, rated under CSA A440.2 or NFRC procedures, with air leakage at or below 1.5 L/s/m², and an effective date of 1 January 2029.
Achieved whenthe final Version 6.0 is published and Canada's list of certified products shows windows certified to it.
Claimed by
Natural Resources Canada, July 2026
Open
01
Read the safety-glass mark
A phone app and an open photo set that find and read the small permanent mark etched, sandblasted or fired into the corner of tempered and laminated glass, decode what it claims (the maker, the standard and the category, such as "CPSC 16 CFR 1201 CAT II"), and look up any certification code it carries in the Safety Glazing Certification Council's public directory of certified products. The work is photographs only, of glass already installed in doors, shower enclosures and panels beside doors.
The briefCredit · unclaimed02
Pooled field water-test results
A shared, anonymized table of field water tests of installed windows and doors (ASTM E1105 or AAMA 502), giving for each test the product type, its rating, the test pressure, pass or fail, where the water came in (through the window itself or at its joint with the wall) and what was fixed. Fill it from reports that public building owners hold and from consultants willing to share theirs.
The briefCredit · unclaimed
Open 01 · the brief
Read the safety-glass mark
The first to publish is credited on this line, and the deliverable moves from Open to Achieved.
What to build
A phone app and an open photo set that find and read the small permanent mark etched, sandblasted or fired into the corner of tempered and laminated glass, decode what it claims (the maker, the standard and the category, such as "CPSC 16 CFR 1201 CAT II"), and look up any certification code it carries in the Safety Glazing Certification Council's public directory of certified products. The work is photographs only, of glass already installed in doors, shower enclosures and panels beside doors.
Why it is still open
Once glass is installed, its mark is the usual way an inspector or buyer tells safety glass from ordinary glass, but it is small, sits in a corner and can be hidden by the frame or worn away, by a window installer's guide. The National Glass Association's 2022 technical paper on marking says the details on such a permanent label are "often coded or abbreviated", with the details of fabricators in the voluntary certification program kept on the Safety Glazing Certification Council's website. Searches of the web, app stores and GitHub in November 2026 found glass-measuring apps, quality-control counters and screen-protector finders, but no tool that reads these marks or looks them up.
Built with what exists
Phone cameras with a clip-on macro lens; CVAT for labeling; an open text detection and recognition model such as PaddleOCR; a small vision-language model to turn each mark into fields, checked by a person; the SGCC certified products directory.
First step
Photograph the mark on every pane in the doors and showers of ten homes, transcribe them, and match the codes to the SGCC directory by hand.
Done looks like
An open set of at least 2,000 photographed marks from at least 100 makers, each transcribed by two people, and a reader that gets the standard and category right on at least 95% of marks photographed in homes it was not trained on, with every error listed.
Open 02 · the brief
Pooled field water-test results
The first to publish is credited on this line, and the deliverable moves from Open to Achieved.
What to build
A shared, anonymized table of field water tests of installed windows and doors (ASTM E1105 or AAMA 502), giving for each test the product type, its rating, the test pressure, pass or fail, where the water came in (through the window itself or at its joint with the wall) and what was fixed. Fill it from reports that public building owners hold and from consultants willing to share theirs.
Why it is still open
Beers and Smith found in 1999 that most windows in new construction failed their first field test but published no shared data. A 2022 IIBEC Interface article says reliable failure rates "are likely not available", because few projects test all their windows, so testing plans rest on what the project team is willing to accept. GitHub searches in November 2026 for "water penetration test" and "window leak test" found nothing relevant.
Built with what exists
Public-records requests to school districts, universities and state building agencies for envelope commissioning and test reports; a low-cost AI model that reads test-report PDFs into rows, checked by a person (see the Note "The jobs agents repeat"); Python, with the sample-size method of ASTM E122 to put confidence ranges on failure rates.
First step
Send public-records requests for the envelope test reports of ten recently finished public buildings, and code the first five reports by hand.
Done looks like
At least 1,000 field tests from at least 50 projects in a public table, a second person checking 100 rows against the reports, and published failure rates by product type and by where the water came in, with confidence ranges.